<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Khushwant Sehgal on Medium]]></title>
        <description><![CDATA[Stories by Khushwant Sehgal on Medium]]></description>
        <link>https://medium.com/@khushwantsehgal?source=rss-aaca1df59fb8------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*B7y-42ieQfI41HQd.png</url>
            <title>Stories by Khushwant Sehgal on Medium</title>
            <link>https://medium.com/@khushwantsehgal?source=rss-aaca1df59fb8------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 20:40:46 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@khushwantsehgal/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[7 Ways Conversational AI Fails Silently, And How to Measure Them]]></title>
            <link>https://medium.com/@khushwantsehgal/7-ways-conversational-ai-fails-silently-and-how-to-measure-them-7335305b778f?source=rss-aaca1df59fb8------2</link>
            <guid isPermaLink="false">https://medium.com/p/7335305b778f</guid>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[claude-code]]></category>
            <category><![CDATA[conversational-ai]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[Khushwant Sehgal]]></dc:creator>
            <pubDate>Tue, 24 Mar 2026 11:29:13 GMT</pubDate>
            <atom:updated>2026-03-24T11:29:13.727Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A practical framework for evaluating conversational AI quality beyond pass/fail, built inside Adobe, applicable to any AI agent or LLM-powered interface you’re building.</em></p><p>Adobe has been betting big on generative AI and conversational interfaces across its enterprise products. One of the newest additions is the chat-style assistant in AEM Forms, a conversational AI that lets non-technical users describe a form in plain English and watch it materialize in real time.</p><p>But shipping a generative AI feature is the easy part. Evaluating whether the <em>experience</em> of using it is any good — that’s harder. Most teams have no framework for measuring conversational AI quality beyond whether the output was technically correct.</p><p>That’s what <strong>Formsight</strong> was built to answer. It’s an internal AI evaluation pipeline that takes any form from a user’s AEM environment, reverse-engineers its structure into the exact prompts a human would type, executes those prompts against a live AI session, and grades every interaction across seven UX dimensions — producing a repeatable, deterministic AI UX quality scorecard.</p><p>More importantly, the seven dimensions it scores aren’t specific to form builders. They apply to any LLM-powered agent or conversational interface where a user types a request and expects action.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/966/1*JbZ_nxKcSxPf-7wAEv6cBw.png" /><figcaption><em>A real Formsight scorecard. Trust &amp; Confidence and Error Recovery fail; Performance and Contextual Guidance pass. The weighted score surfaces what raw pass/fail never would</em></figcaption></figure><blockquote>Thanks to Claude Code, the first version was ready in under four days.</blockquote><h3>Why Conversational AI Evaluation Goes Beyond Pass/Fail</h3><p>Most QA stops at pass/fail. A field either exists or it doesn’t.</p><p>But when your product is a conversational AI agent, pass/fail misses the point entirely. The user didn’t click a button,they had a conversation. They described what they wanted, watched the AI respond, and made an unconscious judgment about whether to trust it enough to keep going. That judgment happens in the space between prompt and response.</p><p>The seven AI UX dimensions below are designed to measure exactly that space.</p><h3>7 UX Dimensions for Evaluating Conversational AI Quality</h3><h3>1. Explainability: “What Did You Just Do?”</h3><p>This is the most fundamental question a user asks after any AI action. In HCI literature it maps to <a href="https://www.nngroup.com/articles/ten-usability-heuristics/">system visibility</a>,<a href="https://www.cs.umd.edu/~ben/goldenrules.html">Shneiderman’s first principle of interface design</a> and a cornerstone of <a href="https://www.nngroup.com/articles/ten-usability-heuristics/">Nielsen’s heuristic evaluation</a>.</p><p>An AI that does the right thing <em>silently</em> is almost as bad as one that does the wrong thing loudly. Users need the AI to narrate its work.</p><p><strong>In Formsight:</strong> When a user says “add correspondence address panel to the contact section,” the AI might succeed perfectly and say nothing. No summary, no field count, no confirmation. Formsight checks whether responses contain creation summaries, field counts, and action descriptions. A 20-character reply to a prompt that specified 12 fields with validation rules is flagged as a UX failure,even if every field was created correctly,because the user has no way to verify the result without manually counting.</p><blockquote><strong><em>In your agent:</em></strong><em> Ask yourself: </em>after my agent acts, can the user tell what happened without checking for themselves?<em> Whether your agent writes an email, exports data to Salesforce, or creates a document,the response should summarize what was done (and what wasn’t). Even a brief “Exported 47 contacts to Salesforce,3 records skipped due to missing email fields” is worlds better than “Done.”</em></blockquote><h3>2. Trust &amp; Confidence: “Can I Believe You?”</h3><p>Trust is asymmetric. It takes dozens of correct interactions to build and a single unexplained discrepancy to break. In conversational UIs that lead to action, trust breaks most often through what Formsight calls <strong>silent drops</strong>,the user asks for 10 fields, the AI creates 7, and the response makes no mention of the missing three.</p><p>A trustworthy system doesn’t just succeed or fail; it accounts for the gap between what was requested and what was delivered.</p><p><strong>In Formsight:</strong> “I created 7 of your 10 fields. I couldn’t create the file upload, date range, or signature fields because they require manual configuration” builds trust even though it reports partial failure.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MEJS4gNS4VQ95SdumC6DsQ.png" /></figure><p><em>Same outcome,7 fields created,but only one of these responses builds trust.</em> Formsight detects count mismatches between prompts and responses, missing uncertainty language on ambiguous input, and operator mismatches in conditional rules (a dropdown that should use “equals” being configured with “contains”).</p><blockquote><strong><em>In your agent:</em></strong><em> Never let your agent go silent about partial completion. If it couldn’t do something, it should say so,and say why. That transparency is what earns the user’s willingness to try again.</em></blockquote><h3>3. Conversational Intelligence: “Are You Actually Listening?”</h3><p>A good conversational partner adapts to what the user actually needs. Formsight evaluates two sides of this:</p><p><strong>Context-appropriate confirmation.</strong> When a user provides a highly specific, unambiguous prompt,five field names, types, and validation rules,the AI should act, not ask “Would you like me to proceed?” That wastes time and signals the AI didn’t register the specificity. In <a href="https://www.sas.upenn.edu/~haroldfs/dravling/grice.html">Grice’s conversational maxims</a>, this violates the maxim of quantity.</p><p>That said, <strong>human-in-the-loop confirmation is valuable</strong> when the action is irreversible or high-stakes. The right design is to make confirmation behavior configurable,as Cursor, ChatGPT, and most mature agent products do via user settings. The failure mode Formsight flags isn’t confirmation itself; it’s confirmation triggered by prompts that leave no room for ambiguity.</p><p><strong>Under-clarification.</strong> When a user types something genuinely vague,”make me a form” with no field names, no context, no constraints,a good AI asks a follow-up. A bad one guesses and builds something generic.</p><p><em>How Formsight detects this:</em> Prompts under 80 characters with no field specifications (no mention of field names, types, or counts) are considered vague. If the AI processed one of these without asking a clarifying question, that’s a miss,not because short prompts are always bad, but because a prompt that short almost certainly lacks the specificity needed to build anything meaningful. The 80-character threshold is calibrated to the domain: a form creation request needs at minimum a field type and label to be actionable.</p><p><strong>Partial processing.</strong> The subtlest failure: the user gives five instructions in one message, and the AI addresses three with no acknowledgment of the other two. This is what HCI researchers call a <a href="https://www.interaction-design.org/literature/book/the-glossary-of-human-computer-interaction/gulf-of-evaluation-and-gulf-of-action">gulf of evaluation</a>,the user can’t tell whether the AI ignored the instructions, failed to understand them, or plans to address them later.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aXZp4OUpRWwJ4W40o4FAew.png" /></figure><p><em>The gulf of evaluation: the user can observe the output but cannot determine the AI’s intent or completeness.</em></p><p><em>Quick fix for your agent:</em> Explicitly handle compound prompts,either process all instructions and confirm each, or respond with “I handled X, Y, and Z; steps ④ and ⑤ will require a follow-up.”</p><blockquote><strong><em>In your agent:</em></strong><em> Build a vagueness detector. If input falls below a complexity threshold for your domain, ask before acting. And if your agent ever processes only part of a multi-step instruction, it must say so.</em></blockquote><h3>4. Error Recovery: “Something Went Wrong,Now What?”</h3><p>Every system fails. The question is whether the failure is a dead end or a detour.</p><p>Formsight checks three things in error responses:</p><ul><li><strong>Specificity:</strong> Does the error explain what actually went wrong, or does it default to “something went wrong, please try again”? Generic errors are the conversational equivalent of a shrug.</li><li><strong>Actionability:</strong> Does the response suggest a recovery path? Words like “try,” “instead,” and “alternatively” signal the AI is helping the user navigate around the failure.</li><li><strong>Professionalism:</strong> Did technical internals leak into the user-facing message? Stack traces, CSS selectors, and internal error codes in a conversational response destroy the illusion that the user is talking to an intelligent assistant.</li></ul><p>Formsight also flags <strong>transient unreliability</strong> (same prompt fails, then succeeds on retry) and <strong>category-specific failure spikes</strong> (one operation type failing at a disproportionately higher rate). Both patterns are invisible to individual users but reveal systemic issues at scale.</p><blockquote><strong><em>In your agent:</em></strong><em> Write your error messages last,after you’ve thought about what the user will actually do next. Every error should answer: “What went wrong, and what can I do about it?”</em></blockquote><h3>5. Response Quality: “Did You Phone That In?”</h3><p>Not every response needs to be a paragraph. But every response needs to be proportionate to the request.</p><p><strong>In Formsight:</strong> A 200-character prompt specifying field types, labels, and validation rules deserves more than a 15-character “Done!” Formsight measures the ratio between prompt complexity and response length, catches <strong>echo-back</strong> (responses where &gt;80% of words are repeated from the user’s prompt, adding nothing), and flags <strong>truncation</strong> (responses ending mid-sentence, suggesting the output was cut off).</p><blockquote><strong><em>In your agent:</em></strong><em> Set a proportionality rule: the more the user specified, the more your agent should confirm. A brief action deserves a brief confirmation. A complex operation deserves a structured summary.</em></blockquote><h3>6. Performance: “Why Is This Getting Slower?”</h3><p>Latency in a conversational interface isn’t just a technical metric,it’s a social one. <a href="https://dl.acm.org/doi/10.1145/3290605.3300830">Research on human conversation</a> shows that response delays beyond 3–4 seconds trigger anxiety and reduced trust. AI interfaces get more leeway, but not unlimited.</p><p><strong>Formsight’s performance thresholds:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zRHKx_Hr8pAZhQilczBmuA.png" /></figure><p>Session degradation (second half of a session significantly slower than the first) is a sign of accumulating context or resource exhaustion,something users rarely articulate but always feel.</p><blockquote><strong><em>In your agent:</em></strong><em> Instrument your p50 and p90 latency separately. p50 is your median,what a typical user experiences. p90 is the slowest 1-in-10 requests, the tail that users actually complain about. Your p50 can look healthy while your p90 is painful. And if your sessions degrade over time, investigate context accumulation before optimizing model calls.</em></blockquote><h3>7. Contextual Guidance: “What Should I Do Next?”</h3><p>The best conversational interfaces don’t just respond,they guide. After creating a form skeleton, does the AI suggest adding fields? After configuring validation, does it mention conditional visibility rules?</p><p>This is the least weighted dimension,it’s a nice-to-have, not a must-have. But its presence separates an AI that functions from an AI that collaborates.</p><blockquote><strong><em>In your agent:</em></strong><em> After completing a task, consider one proactive next step. Not a menu of options,just one logical continuation that moves the user forward.</em></blockquote><h3>The AI UX Quality Scorecard: Weighted Dimensions</h3><p>Not all seven dimensions carry equal weight. Formsight’s scoring reflects a hierarchy of user needs:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ndHkXycpIHW6JK3gu1xkKQ.png" /></figure><p>The hierarchy mirrors Maslow: don’t optimize for delight until you’ve secured the fundamentals. <strong>Pass/Warn/Fail</strong> thresholds are set per dimension based on empirical patterns from hundreds of sessions,e.g., a trust score below 6/10 indicates measurable count mismatches or missing uncertainty language; below 4/10 indicates systematic silent drops.</p><h3>How Formsight Evaluates Conversational AI: The Pipeline</h3><p>The pipeline: analyze a form’s JSON model from the user’s AEM environment (org-specific forms, not public templates) → extract ground truth structure → generate the natural-language prompts a human would type to recreate it → execute those prompts against a live authenticated browser session → grade every AI response with the seven-dimension scorecard.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0G2Ivjn8REWUvhoX2yvZyA.png" /><figcaption><em>Formsight’s five-step pipeline. Every stage is automated,from reading the form structure to typing prompts in a real browser to grading the AI’s responses.</em></figcaption></figure><p>The entire evaluation layer is <strong>deterministic</strong>,no LLM evaluating another LLM. Just pattern matching and heuristics running in milliseconds. This makes results reproducible, fast, and free of the meta-problem of “who evaluates the evaluator?”</p><h3>Why AI UX Evaluation Matters at Scale</h3><p>The gap between “we shipped a generative AI feature” and “the AI agent works reliably at scale” is enormous,and most teams have no framework to measure it. A 5% silent field-drop rate doesn’t show up in a demo. It shows up when 10,000 enterprise customers are using the AI in production.</p><p>The seven-dimension AI UX scorecard isn’t just a testing framework. It’s a <strong>design specification</strong> for what a good conversational AI agent should feel like,applicable whether you’re building form generators, email agents, CRM integrations, or any LLM-powered interface where a user types a request and trusts the system to act on it.</p><h3>Closing the Loop: From AI Evaluation to Product Improvement</h3><p>The real value isn’t the scorecard,it’s what it enables. Formsight identifies exactly where the AI struggles: low explainability on complex forms with 15+ fields, trust failures when conditional rules involve checkbox operators, performance degradation in sessions exceeding 20 prompts.</p><p>These findings are specific, reproducible, and directly actionable. Each batch run becomes a before-and-after measurement: did the latest model update improve explainability on complex forms?</p><p>It’s the kind of internal tooling that rarely makes headlines but quietly determines whether an AI feature ships as a gimmick or as something people actually rely on.</p><p>Many of the issues and improvement suggestions identified through Formsight are already being considered and incorporated by the development team,closing the loop between measurement and the product itself.</p><p><em>Formsight is an internal Adobe tool, not available as a public product. We’re working on additional dimensions, the existing seven have already surfaced 20+ prioritized issues. More on those in the next post.</em></p><p><em>Formsight was </em><a href="https://medium.com/u/ae5d5a5d0405"><em>Salil Taneja</em></a><em>’s idea, the seed that started it.</em></p><p><em>Anurag Sharma pushed us to go deeper on UX dimensions, which shaped the direction of this work significantly.</em></p><p><em>The seven dimensions themselves owe a great deal to </em><a href="https://medium.com/u/acf05f9fba40"><em>Utkarsha Sharma</em></a><em>, designer and collaborator, who helped brainstorm, pressure-test, and clarify the concepts around UX in conversational interfaces, her inputs are woven into every dimension described here.</em></p><p><em>Thanks also to </em><a href="https://medium.com/u/4c58e161d4d2"><em>Shivani Batra</em></a><em> and </em><a href="https://medium.com/u/405e410b72eb"><em>Udayanmishra</em></a><em> for thorough review and feedback on this blog.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7335305b778f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Built a On-Device Self-Correcting Experience Without Overusing]]></title>
            <link>https://medium.com/@khushwantsehgal/how-to-build-a-self-correcting-ux-system-without-overusing-llms-fcd17f639123?source=rss-aaca1df59fb8------2</link>
            <guid isPermaLink="false">https://medium.com/p/fcd17f639123</guid>
            <category><![CDATA[personalization]]></category>
            <category><![CDATA[user-experience]]></category>
            <category><![CDATA[mobile-development]]></category>
            <category><![CDATA[product-development]]></category>
            <category><![CDATA[software-engineering]]></category>
            <dc:creator><![CDATA[Khushwant Sehgal]]></dc:creator>
            <pubDate>Mon, 23 Mar 2026 16:52:03 GMT</pubDate>
            <atom:updated>2026-03-24T03:56:02.349Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*tF8lzXV2s9Tfltwev82iHg.png" /><figcaption><em>A high-level view of the self-learning loop: from error correction to automated success.</em></figcaption></figure><h3>I Built a On-Device Self-Correcting Experience Without Overusing LLMs… So Should You</h3><p>There’s a specific kind of digital fatigue that defines the modern app experience: correcting the same mistake twice. You open your finance app, see a cryptic charge like “AMZN-MKT-UK-9921”, manually categorize it as Office Supplies, and move on. But then the same charge appears next month, and there it is: Uncategorized, waiting for the same manual fix.</p><p>This friction is the enemy of User Experience (UX). It is why hyper-personalization and adaptive UI are becoming the gold standards of modern product design. One of the underlying patterns behind these trends is the correction-memory loop.</p><h3>Automation is only as useful as its memory. True personalization begins when the app stops asking the same questions.</h3><h3>What is a correction-memory loop?</h3><p>In practice, a correction-memory loop is a user feedback loop where manual fixes are treated as a highly reliable source of truth in that context. Instead of relying on static rules, the system remembers the user’s explicit intent and reuses it to adapt future behavior.</p><p>Think of it like a smart roommate.</p><ul><li>The Average App: Keeps making the same mistake no matter how many times you correct it.</li><li>The Adaptive App: Remembers you only take one sugar. From that day on, it just works.</li></ul><h3>How to Build It</h3><p>Building a correction-memory system doesn’t require a PhD in AI. It requires a disciplined approach to three architectural pillars. If you’re a developer or student, here is the blueprint you need to follow:</p><h4>1. The Contextual Identifier (The Fingerprint)</h4><p>You need a way to consistently identify repeatable inputs without storing massive amounts of text. We use hashing.</p><ul><li>Take the raw, unformatted input (e.g., a raw OCR text block or a cryptic merchant string).</li><li>Run it through a hashing algorithm like SHA-256 (or a lighter option like MurmurHash if cryptographic strength isn’t needed). This creates a compact, effectively unique fingerprint<strong><em> </em></strong>for that normalized input.</li><li>Normalize your data (remove extra spaces/case) before hashing to increase the hit rate.</li></ul><p>This works well for exact matches, but won’t capture variations of the same input. Slight changes (like different merchant formats) will produce different fingerprints. This can be improved with stronger normalization or fuzzy matching (e.g., token matching or string similarity), but it’s an important constraint to be aware of.</p><h4>2. The Persistent Memory (The Repository)</h4><p>You need a local brain to store the mapping between the Fingerprint and the user’s correction.</p><ul><li>Create a simple database table or key-value store.</li><li>Set schema, like input_hash (Primary Key) | user_correction_fields (JSON or separate columns).</li><li>When to save? Record the entry only at the Point of Commitment (e.g., when the user taps “Confirm” or “Save”).</li></ul><p>In practice, this should usually be scoped per user to avoid conflicts across different behaviors.</p><h4>3. The Interception Layer (The Injector)</h4><p>This is where the system actually becomes adaptive. You must intercept the data pipeline before it reaches the UI.</p><pre>def process_data(raw_input):<br> # Step A: Identify the input<br> fingerprint = generate_hash(normalize(raw_input))<br> <br> # Step B: Check Memory<br> previous_correction = database.get_correction(fingerprint)<br> <br> if previous_correction:<br> # Use the stored correction immediately <br> return previous_correction<br> else:<br> # Fall back to default automation/guessing<br> return default_recognition_logic(raw_input)</pre><p>This is not a prediction layer. The system is simply reusing a previously confirmed answer.</p><h3>An Example of High-Efficiency Personalization</h3><p>Hashing allows us to condense complex, normalized data into compact, effectively unique fingerprints for instant recall.</p><p>Imagine an expense tracker called “SpendWise”. When you import a transaction string, the app generates a fingerprint. If you manually link “XYZ-CLOUD-SRV” to “Hosting Fees,” the app “commits” that mapping to memory.</p><p>The next time that fingerprint appears, the app bypasses the guessing phase and applies the user’s preference in milliseconds. This approach is instant, can run fully on-device, and provides real-time personalization without expensive API fees.</p><h3>The Teach, Confirm, Re-Scan Protocol</h3><p>How do you verify your loop is closed? Use this simple stress test:</p><ol><li>Identify a data point where the system predictably fails or provides a generic result.</li><li>Manually provide the correct data. This is the correction phase.</li><li>Perform the action again and verify the system reuses the stored correction.</li></ol><p>If the system reuses your previous fix correctly, you’ve successfully implemented a truly adaptive user experience.</p><h3>The Bottom Line</h3><p>Modern software engineering is moving away from one-size-fits-all automation. By implementing a correction-memory loop, you’re not just fixing a bug. You’re reducing repeated user effort every time the same situation occurs.</p><p><strong>The future of UX isn’t about more features; it’s about less friction through persistent memory.</strong></p><p>Have you built a similar loop in your own stack? Or are you still fighting the ‘three-sugar’ roommate? Let’s discuss in the comments.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fcd17f639123" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Claude Code Login Screen Issue with Amazon Bedrock in Cursor/VS Code]]></title>
            <link>https://medium.com/@khushwantsehgal/setting-up-claude-code-extension-in-cursor-vs-code-with-amazon-bedrock-a871af187045?source=rss-aaca1df59fb8------2</link>
            <guid isPermaLink="false">https://medium.com/p/a871af187045</guid>
            <category><![CDATA[claude-code]]></category>
            <category><![CDATA[cursor]]></category>
            <category><![CDATA[amazon]]></category>
            <dc:creator><![CDATA[Khushwant Sehgal]]></dc:creator>
            <pubDate>Thu, 12 Feb 2026 15:01:32 GMT</pubDate>
            <atom:updated>2026-02-12T15:17:49.761Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*12D0fbETmVEYjOn0iCuVuA.png" /></figure><p>I recently tried to set up the Claude Code extension in Cursor IDE using Amazon Bedrock, and ran into the annoying login screen issue. After some back and forth, I finally got it working. Here’s what I did.</p><p>I wanted to use Claude Code in Cursor, but I have AWS Bedrock access, not an Anthropic API key. The documentation wasn’t super clear on how to set this up, and I kept hitting a login screen even after installing the extension.</p><h3>What You’ll Need</h3><ul><li>Cursor IDE installed</li><li>Amazon Bedrock API key (<a href="https://aws.amazon.com/blogs/machine-learning/accelerate-ai-development-with-amazon-bedrock-api-keys/">get one here</a>)</li><li>Your AWS region (like us-east-1)</li></ul><h3>Create the Configuration File</h3><p>First, I created a configuration file at ~/.claude/settings.json:</p><pre>mkdir -p ~/.claude</pre><p>Then opened the file in my editor and added:</p><pre>{<br>  &quot;$schema&quot;: &quot;https://json.schemastore.org/claude-code-settings.json&quot;,<br>  &quot;env&quot;: {<br>    &quot;CLAUDE_CODE_USE_BEDROCK&quot;: &quot;1&quot;,<br>    &quot;AWS_REGION&quot;: &quot;us-east-1&quot;,<br>    &quot;AWS_BEARER_TOKEN_BEDROCK&quot;: &quot;YOUR_BEDROCK_API_KEY_HERE&quot;<br>  }<br>}</pre><p><strong>Important:</strong> Replace YOUR_BEDROCK_API_KEY_HERE with your actual key, and update AWS_REGION if needed.</p><h3>Install the Extension in Cursor</h3><ol><li>Opened Cursor</li><li>Pressed Cmd+Shift+X (Mac) or Ctrl+Shift+X (Windows/Linux)</li><li>Searched for “Claude Code”</li><li>Clicked Install</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Pk09T2BgLxhLVbgs7uYfjg.png" /></figure><h3>The Critical Step — Disable Login Prompt</h3><p><strong>This is where I got stuck.</strong> Even with everything configured, I kept seeing the login screen. The fix was to disable the login prompt in Cursor settings.</p><h3>Here’s what worked for me:</h3><ol><li>Pressed Cmd+Shift+P (or Ctrl+Shift+P)</li><li>Typed: Preferences: Open Settings (UI)</li><li>Searched for: claudeCode.disableLoginPrompt</li><li><strong>Checked the box</strong> for “Disable Login Prompt”</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/796/1*j715qEqnsMlRVwfO4rCrcg.png" /></figure><h3>Alternative Method (Edit JSON Directly):</h3><p>If you prefer editing JSON:</p><ol><li>Press Cmd+Shift+P → Preferences: Open User Settings (JSON)</li><li>Add this line:</li></ol><pre>&quot;claudeCode.disableLoginPrompt&quot;: true</pre><h3>Restart Cursor</h3><p><strong>Completely quit</strong> Cursor (not just reload) and reopen it.</p><ul><li>Mac: Cmd+Q then reopen</li><li>Windows/Linux: Close all windows and reopen</li></ul><h3>Step 5: Test It</h3><ol><li>Opened any file in Cursor</li><li>Clicked the spark icon in the top-right corner</li><li>Sent a test message</li></ol><p>![Screenshot: Claude Code panel working in Cursor]</p><p><strong>It worked!</strong> No login screen.</p><h3>Troubleshooting</h3><h3>Still Seeing the Login Screen?</h3><ul><li>Double-check that claudeCode.disableLoginPrompt is set to true</li><li>Make sure you <strong>completely restarted</strong> Cursor (not just reloaded the window)</li><li>Verify the setting is in Cursor’s settings, not just your project settings</li></ul><h3>Configuration Not Working?</h3><ul><li>Check that ~/.claude/settings.json has valid JSON syntax (no trailing commas!)</li><li>Verify your Bedrock API key is correct</li><li>Make sure you have Bedrock access in the AWS region you specified</li></ul><h3>Quick FAQs</h3><p><strong>Q: Do I need an Anthropic API key?</strong></p><p>A: Nope! That’s the whole point of using Bedrock.</p><p><strong>Q: What AWS regions work?</strong><br>A: Common ones are us-east-1, us-west-2, eu-west-1. Check which regions have Claude models enabled in your Bedrock console.</p><p><strong>Q: Can I use this setup in the terminal too?</strong></p><p>A: Yes! The CLI uses the same ~/.claude/settings.json file.</p><h3>Summary</h3><p>Here’s the quick version:</p><ol><li>Create ~/.claude/settings.json with Bedrock config</li><li>Install Claude Code extension in Cursor</li><li><strong>Enable “Disable Login Prompt” in Cursor settings</strong> (this is key!)</li><li>Completely restart Cursor</li><li>Open the Claude Code panel and test</li></ol><p>The biggest gotcha is Step 3 — you MUST disable the login prompt in Cursor’s settings, or you’ll keep seeing the login screen even with everything configured correctly.</p><p>Hope this helps someone else who was stuck on the same issue!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a871af187045" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[8 Lessons from Building AI Agents]]></title>
            <link>https://medium.com/@khushwantsehgal/8-lessons-from-building-ai-agents-2b9754f2af79?source=rss-aaca1df59fb8------2</link>
            <guid isPermaLink="false">https://medium.com/p/2b9754f2af79</guid>
            <category><![CDATA[openapi-specification]]></category>
            <category><![CDATA[enterprise-ai-workflows]]></category>
            <category><![CDATA[ai-agent]]></category>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[enterprise]]></category>
            <dc:creator><![CDATA[Khushwant Sehgal]]></dc:creator>
            <pubDate>Fri, 29 Aug 2025 06:27:35 GMT</pubDate>
            <atom:updated>2025-09-04T13:33:36.512Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UhVw7UnRdWSAATxBav2MSg.jpeg" /></figure><p>When most Vibe Coders talk about AI agents, the conversation often revolves around prompts and large language models (LLMs). But in practice, building AI agents that deliver real business value — especially in enterprise environments — requires more than clever prompting. Architecture, interoperability, analytics, and documentation practices determine whether an agent remains a demo or evolves into a product that drives adoption.</p><p>Recently, I worked on two AI-Agents. Here are 8 key lessons I learnt — applicable to anyone designing enterprise-ready AI agents.</p><h3>1. Keep frontend and backend separate</h3><p>A common mistake is tying the agent frontend and backend too closely together. By building the backend independently, I avoided dependency chains that slow development. Once the backend was stable, I layered the frontend on top. This modular AI agent architecture makes scaling, testing, and iteration far smoother. By keeping the backend AI logic independent of the frontend UI, I could optimize reasoning, monitor token usage, and scale LLM calls without slowing down the user interface.</p><h3>2. <strong>Support open standards and protocols</strong></h3><p>In enterprise environments, interoperability is critical. Supporting open protocols such as Google’s Agent2Agent (A2A) and the Model Context Protocol (MCP) ensures agents can communicate and collaborate effectively across platforms. A2A enables secure agent-to-agent interactions, while MCP standardizes how agents access external tools and resources. Using these open standards reduces vendor lock-in and enhances long-term adoption by allowing agents to integrate seamlessly with other enterprise systems.</p><h3>3. Stay LLM-agnostic</h3><p>The LLM landscape is evolving rapidly, and enterprises face unique compliance, performance, and cost considerations. Designing agents to be LLM-agnostic allows switching between models without rewriting core logic. This flexibility helps protect against vendor lock-in and adapts to changes in the AI ecosystem.</p><h3>4. Capture analytics from day one</h3><p>Without usage data, you’re flying blind. From the very first deployment, I would instrument our agents to collect <strong>detailed usage analytics</strong> — tracking adoption, performance, and edge cases. This data would provide actionable insights for continuous improvement and provided measurable ROI for stakeholders.</p><h3>5. Build documentation in parallel with development</h3><p>For agents to succeed in enterprise adoption, documentation automation must be treated as part of the build process. Writing public documentation alongside development kept everything accurate and up-to-date, making onboarding faster and reducing support overhead. Documentation isn’t an afterthought — it’s a product feature. Keeping documentation ready helped me easily deploy Agents on production environments.</p><h3>6. Maintain a living OpenAPI specification</h3><p>We treated our OpenAPI specification as the single source of truth for backend capabilities. By updating it continuously, we enabled smoother frontend development, automated client generation, and easier integration testing. In enterprise automation, a living API spec accelerates both development and adoption.</p><h3>7. Use Nudge Architecture to drive adoption</h3><p>Technology alone doesn’t guarantee usage. In Forum Scout, we applied Nudge Architecture — embedding subtle behavioral cues into the workflow to encourage the right actions at the right time. Small nudges (like timely reminders, prioritization hints, or contextual suggestions) helped users engage with the agent more naturally, improving adoption and sustained usage.</p><p>For enterprise AI agents, combining strong technical foundations with behavioral nudges can bridge the gap between capability and actual user engagement.</p><h3>8. Think adoption, not just demos</h3><p>The final and most important lesson: flashy demos don’t guarantee adoption. What makes an AI agent <em>usable</em> in the enterprise is robustness — clean architecture, protocol interoperability, analytics, documentation, and nudges. These foundations turn proof-of-concepts into scalable automation tools.</p><h3>Final Thoughts</h3><p>The projects started as experimental AI agents, but the lessons I learnt highlight a universal truth: building enterprise-ready AI automation is about more than the LLM.</p><p>If you’re designing AI agents for enterprise workflows, focus not only on intelligence but also on architecture, standards, documentation, analytics, and nudges. That’s how you build AI systems that move from concept to adoption.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2b9754f2af79" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Connect Any Data, Anywhere: How AEM Forms Powers Your Enterprise]]></title>
            <link>https://medium.com/@khushwantsehgal/connect-any-data-anywhere-how-aem-forms-powers-your-enterprise-a2c337f670ec?source=rss-aaca1df59fb8------2</link>
            <guid isPermaLink="false">https://medium.com/p/a2c337f670ec</guid>
            <category><![CDATA[management]]></category>
            <category><![CDATA[productivity]]></category>
            <category><![CDATA[forms]]></category>
            <category><![CDATA[integration]]></category>
            <category><![CDATA[customer-experience]]></category>
            <dc:creator><![CDATA[Khushwant Sehgal]]></dc:creator>
            <pubDate>Mon, 02 Jun 2025 19:02:56 GMT</pubDate>
            <atom:updated>2025-06-03T04:53:48.901Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tZR8NluJGQyFHaiWE2HnWQ.png" /></figure><h3><strong>Tired of Data Silos? How Connected Forms are Revolutionizing Enterprise Workflows</strong></h3><p>You have powerful systems running your business — CRMs, databases, workflow engines. Your forms are where critical interactions happen. So, how do we bridge that gap seamlessly?</p><p>That’s where AEM Forms comes in. Adobe has designed AEM Forms to be your central <strong>integration hub</strong>, seamlessly orchestrating data flow between your forms and virtually any backend system. <strong>It’s not about migrating data, but about making your existing systems work together, intelligently, through your forms.</strong></p><p><strong>Let’s See it in Action: The Modern Loan Application</strong></p><p>Imagine a customer, Sarah, applying for a loan on your website.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8x-VEUgitIczSnXyz_gbaA.png" /></figure><p><strong>A Smart Start (Prefill — Fan-In): </strong>As Sarah begins, AEM Forms is already working:</p><ol><li>It pulls her profile from your CRM (like Salesforce).</li><li>Checks her eligibility via a credit score API.</li><li>Fetches current rates from your core banking system (via JDBC).<br>Sarah gets a personalized, partly-filled form. Effortless for her, efficient for you.</li></ol><p><strong>Data in Motion (Submission — Fan-Out): </strong>Sarah hits “Submit.” AEM Forms intelligently routes the information:</p><ol><li>The application goes to your Loan Processing System (via REST API).</li><li>An Adobe Sign workflow is triggered for digital signatures.</li><li>A record is stored in Microsoft SharePoint or Azure Blob Storage.</li><li>An alert pops up in Adobe Workfront for the operations team.</li></ol><p>In this single workflow, AEM Forms talked to multiple different systems — retrieving and sending data, all in real-time. <strong>This loan application is a perfect illustration of fan-in for prefill (pulling from CRM, credit API, banking system) and fan-out for submission (sending to Loan Processing, Adobe Sign, SharePoint, Workfront), all orchestrated by AEM Forms.</strong> That’s the connected experience we deliver.</p><h3><strong>How AEM Forms Makes Integration Simple</strong></h3><p>AEM Forms offers multiple ways to connect your forms to your enterprise data:</p><h4><strong>Effortless Integration with Pre-Built Connectors</strong></h4><p>AEM Forms gets you started quickly with a rich set of <strong>out-of-the-box (OOTB)</strong> connectors for many popular enterprise systems. These are your fast track to integration, often configured with minimal effort:</p><ul><li><strong>CRM Systems:</strong> <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/use-form-data-model/configure-msdynamics">Microsoft Dynamics 365</a>, <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/use-form-data-model/aem-forms-salesforce-integration">Salesforce</a></li><li><strong>Cloud Storage &amp; Collaboration:</strong> <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-core-components/create-an-adaptive-form-on-forms-cs/connect-to-sharepoint/configure-submit-action-sharepoint">Microsoft SharePoint</a>, <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-core-components/create-an-adaptive-form-on-forms-cs/configure-submit-action-azure-blob-storage">Azure Blob Storage</a></li><li><strong>Workflow &amp; Productivity:</strong> <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/submit-adaptive-form-to-workfront-fusion">Adobe Workfront</a>, <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-foundation-components/use-adobe-sign/working-with-adobe-sign">Adobe Sign</a>, <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/integrate-adaptive-form-with-market-engage/integrate-form-to-marketo-engage">Adobe Marketo</a>, <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/forms-microsoft-power-automate-integration">Microsoft Power Automate</a></li><li><strong>Customer Data Platforms:</strong> <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/aem-forms-aep-connector">Adobe Experience Platform (AEP)</a> for rich prefill and personalization.</li></ul><p>These connectors integrate seamlessly with <strong>AEM Workflows for powerful automation and orchestration</strong>, and with the Form Data Model for precise data mapping.</p><h4><strong>The Powerhouse: Your ‘Connect-to-Anything’ Engine (Form Data Model)</strong></h4><p>Beyond our pre-built connectors, or for more complex integration scenarios, our <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/use-form-data-model/create-form-data-models"><strong>Form Data Model (FDM)</strong></a> is where the real power lies. It’s an <strong>in-house engine</strong> designed to speak the language of virtually any system using its <strong>intuitive UI. </strong>It talks to:</p><ul><li><strong>RESTful APIs:</strong> The standard for modern web services.</li><li><strong>SOAP Web Services:</strong> Connecting to established enterprise applications.</li><li><strong>JDBC-Compliant Databases:</strong> Direct access to your Oracle, SQL Server, MySQL, PostgreSQL, etc.</li><li><strong>OData Services:</strong> For platforms like SAP and other OData-enabled applications.</li></ul><p>The FDM lets you visually configure these connections, define data structures, and map data between your forms and backend systems — often without writing a single line of code. This is the engine that underpins AEM Forms’ ability to connect to diverse data sources and destinations, <strong>powering both fan-in prefill and fan-out submission scenarios.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FixJhdAf0ZcpA1ViPaAFJw.png" /></figure><h3><strong>The AEM Forms Advantage: Your Data, Your Way</strong></h3><p>With AEM Forms acting as the intelligent layer between your forms and backend systems:</p><ul><li>Enable forms to <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-foundation-components/prepopulate-adaptive-form-fields"><strong>prefill</strong></a> with data pulled from CRM, databases, or other systems, delighting users and ensuring accuracy.</li><li>Allow forms to <strong>submit</strong> data to one or many systems simultaneously, in real-time, via configured connectors.</li><li><strong>Automate complex workflows</strong> triggered by form submissions, integrating with tools like <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-foundation-components/use-adobe-sign/working-with-adobe-sign">Adobe Sign</a> and <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/submit-adaptive-form-to-workfront-fusion">Workfront</a> for seamless process orchestration.</li><li>Significantly <strong>reduce development time</strong> and reliance on custom integration code by leveraging OOTB connectors and the visual FDM.</li><li><strong>Adapt quickly</strong> as your backend systems evolve or new ones are introduced, thanks to the flexible FDM.</li></ul><h3><strong>Ready to Unleash Your Data?</strong></h3><p>If you’re looking to create truly connected form experiences and streamline your data workflows, AEM Forms is ready to help. We believe your forms should be powerful conduits for your business, not roadblocks.</p><p>I’d love to hear about your integration goals. Feel free to reach out, and let’s explore how AEM Forms can empower your enterprise.</p><p><em>What’s your most pressing form integration challenge? </em><a href="https://www.linkedin.com/in/khushwant-singh-sehgal/"><em>Let’s connect!</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a2c337f670ec" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AEM Forms: The Enterprise-Grade Solution for Smart, Scalable, and Secure Form Experiences]]></title>
            <link>https://medium.com/@khushwantsehgal/aem-forms-the-enterprise-grade-solution-for-smart-scalable-and-secure-form-experiences-b63b684923bc?source=rss-aaca1df59fb8------2</link>
            <guid isPermaLink="false">https://medium.com/p/b63b684923bc</guid>
            <category><![CDATA[integration]]></category>
            <category><![CDATA[forms]]></category>
            <category><![CDATA[pdf]]></category>
            <category><![CDATA[aem]]></category>
            <category><![CDATA[data]]></category>
            <dc:creator><![CDATA[Khushwant Sehgal]]></dc:creator>
            <pubDate>Wed, 28 May 2025 13:20:42 GMT</pubDate>
            <atom:updated>2025-05-28T13:20:42.078Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bSorvSPw9ESnHkfgOcUGrw.png" /></figure><p>In an era where digital agility and customer experience define competitive advantage, forms are no longer simple data capture tools — they are critical touchpoints in the user journey. <strong>Adobe Experience Manager (AEM) Forms</strong> is an enterprise-grade solution that transforms how organizations create, manage, and deliver intelligent, responsive, and secure digital forms at scale.</p><p>Whether you’re in government, financial services, healthcare, or retail, AEM Forms empowers teams to build seamless, accessible, and branded form experiences — all while integrating tightly with enterprise systems and automating backend processes.</p><h3>Why Choose AEM Forms?</h3><p>AEM Forms is purpose-built to address the complexities of large-scale form management, offering a comprehensive toolkit that combines <strong>ease of design</strong>, <strong>dynamic behavior</strong>, <strong>data security</strong>, and <strong>workflow automation</strong>.</p><p>Let’s explore its core capabilities.</p><h3>1. Adaptive and Responsive Form Design</h3><p>Creating forms that work across devices isn’t a luxury — it’s a necessity. AEM Forms enables:</p><ul><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-core-components/create-an-adaptive-form-on-forms-cs/creating-adaptive-form-core-components"><strong>Adaptive Forms</strong></a>: Automatically adjust layout and behavior based on device type (desktop, tablet, mobile) and browser.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-core-components/create-an-adaptive-form-on-forms-cs/rule-editor-core-components/rule-editor-core-components"><strong>Visual Form Editor</strong></a>: Drag-and-drop interface with support for reusable fragments, making form creation accessible to non-technical users.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-core-components/using/adaptive-forms/sample-themes-templates-form-data-models-core-components"><strong>Theming and Branding</strong></a>: Customize form styles using CSS to match corporate brand guidelines.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-core-components/create-an-adaptive-form-on-forms-cs/supporting-new-language-localization-core-components"><strong>Multi-Lingual Support</strong></a>: Build and manage forms in multiple languages from a single interface.</li></ul><h3>2. Rich Interactivity with Minimal Coding</h3><p>Forms don’t have to be static. AEM Forms supports:</p><ul><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-core-components/create-an-adaptive-form-on-forms-cs/rule-editor-core-components/rule-editor-core-components-usecases#triggering-multiple-actions-using-the-when-rule"><strong>Conditional Logic</strong></a>: Show or hide fields dynamically based on user input.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-core-components/create-an-adaptive-form-on-forms-cs/rule-editor-core-components/rule-editor-core-components-usecases#streamline-complex-calculations-in-repeatable-panels-with-functions"><strong>Calculated Fields</strong></a>: Perform real-time calculations directly within the form.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-core-components/create-an-adaptive-form-on-forms-cs/create-and-use-custom-function-core-component/custom-function-core-components-use-cases"><strong>Custom Scripting</strong></a>: Use JavaScript to add advanced behavior.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-core-components/create-an-adaptive-form-on-forms-cs/adaptive-form-fragments-core-components"><strong>Form Fragments</strong></a>: Build once, reuse across multiple forms for consistency and efficiency.</li></ul><p>This makes it easy to deliver <strong>personalized</strong>, <strong>intuitive</strong>, and <strong>error-resistant</strong> forms that reduce friction in user journeys.</p><h3>3. Intelligent Data Capture and Validation</h3><p>AEM Forms enhances data accuracy and user satisfaction through:</p><ul><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-foundation-components/prepopulate-adaptive-form-fields"><strong>Pre-filled Forms</strong></a>: Populate fields using user profiles or integrated data sources.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-learn/forms/some-useful-integrations/ocr-data-extraction"><strong>Multi-Channel Capture</strong></a>: Web, mobile, and even scanned paper forms using OCR.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-foundation-components/add-rules-and-use-expressions-in-an-adaptive-form/add-custom-error-handler-adaptive-forms"><strong>Validation Rules</strong></a>: Enforce rules on both client and server sides for robust input validation.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-core-components/using/adaptive-forms/adaptive-forms-components/file-attachment"><strong>Attachments Support</strong></a>: Allow users to upload supporting documents securely.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-65/content/forms/aem-forms-app/work-offline-mode"><strong>Offline Form Filling</strong></a>: The AEM Forms app enables offline data capture with auto-sync.</li></ul><h3>4. Workflow Automation and Task Management</h3><p>Form submission is just the beginning. AEM Forms integrates seamlessly with AEM Workflows to automate downstream processes:</p><ul><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/create-form-centric-workflows/aem-forms-workflow-step-reference"><strong>Form-centric Workflows</strong></a>: Route submissions for review, approval, or further processing.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-learn/foundation/workflow/use-the-workflow-editor"><strong>Visual Workflow Editor</strong></a>: Design complex workflows with stages, conditions, and user assignments.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-learn/foundation/workflow/use-the-inbox"><strong>Task Inbox</strong></a>: A centralized view for employees to manage form-related tasks.</li></ul><p>This capability reduces turnaround times and eliminates manual handoffs in customer or employee journeys.</p><h3>5. Deep Integration Across the Adobe and Enterprise Ecosystem</h3><p>AEM Forms is not an isolated tool — it’s a connected powerhouse.</p><ul><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/adobe-sign-integration-adaptive-forms"><strong>Adobe Sign</strong></a>: Built-in e-signature workflows.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/integrate-aem-forms-with-adobe-analytics"><strong>Adobe Analytics</strong></a>: Track drop-off rates, completion times, and field-level engagement.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-65/content/sites/authoring/aem-adobe-campaign/adobe-campaign-forms"><strong>Adobe Campaign</strong></a><strong> &amp; </strong><a href="https://experienceleague.adobe.com/en/docs/experience-manager-65/content/sites/administering/integration/target"><strong>Target</strong></a>: Personalize form experiences and follow up with relevant communications.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/use-form-data-model/using-form-data-model"><strong>Back-End Integration</strong></a>: Use JDBC, REST/SOAP, LDAP, SharePoint, and more to connect to enterprise systems.</li></ul><h3>6. Document Generation and Personalized Communications</h3><p>Turn form submissions into high-quality documents with:</p><ul><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/adaptive-forms-authoring/authoring-adaptive-forms-foundation-components/generate-document-of-record-for-non-xfa-based-adaptive-forms"><strong>Document of Record (DoR)</strong></a>: Automatically generate a PDF snapshot based on submitted data.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-65/content/forms/getting-started/interactive-communications-overview"><strong>Interactive Communications</strong></a>: Deliver responsive, device-friendly documents.</li><li><strong>Automated Messaging</strong>: Create templated letters, summaries, and statements personalized to each user.</li></ul><p>Ideal for industries that require official records or regulated communication (e.g., finance, insurance, government).</p><h3>7. Enterprise-Grade Security and Compliance</h3><p>AEM Forms ensures that sensitive user data is protected through:</p><ul><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/setup-configure-migrate/forms-groups-privileges-tasks"><strong>Role-Based Access Control</strong>:</a> Define who can view, create, or process forms.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-learn/cloud-service/forms/create-aem-workflow/externalize-workflow"><strong>Data Encryption</strong></a>: In transit and at rest.</li><li><strong>Audit Trails</strong>: Track every interaction with a form.</li><li><strong>Compliance</strong>: Supports GDPR, HIPAA (with proper configuration), and WCAG 2.1 AA accessibility standards.</li></ul><p>This makes AEM Forms a trusted solution for industries with stringent regulatory requirements.</p><h3>8. Flexible Deployment Options</h3><p>Choose the model that aligns with your IT strategy:</p><ul><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-65/content/forms/getting-started/introduction-aem-forms"><strong>On-Premise</strong></a>: Full control over infrastructure and data.</li><li><strong>Managed Services</strong>: Adobe hosted environments.</li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/forms-overview/home"><strong>AEM as a Cloud Service</strong></a>: A scalable, cloud-native SaaS model with continuous updates and elastic scalability.</li></ul><h3>9. Built-In Analytics and Reporting</h3><p>Use <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/integrate-aem-forms-with-adobe-analytics"><strong>Adobe Analytic</strong></a><strong>s</strong> or custom dashboards to monitor:</p><ul><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/view-understand-aem-forms-analytics-reports">Form abandonment rates</a></li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/view-understand-aem-forms-analytics-reports">Average completion times</a></li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/view-understand-aem-forms-analytics-reports">Most/least interacted fields</a></li><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/integrate/services/view-understand-aem-forms-analytics-reports">Workflow efficiency</a></li></ul><p>These insights help teams continuously improve form usability and business outcomes.</p><h3>Unique Differentiators</h3><ul><li><a href="https://experienceleague.adobe.com/en/docs/experience-manager-headless-adaptive-forms/using/overview">Headless architecture</a> for ultimate front-end flexibility and custom form experiences</li><li>AI-powered <a href="https://experienceleague.adobe.com/en/docs/aem-forms-automated-conversion-service/using/introduction">Advanced automated conversion</a> of legacy PDF forms to modern adaptive experiences</li><li>Native support for <strong>Document of Record generation</strong></li><li>Deep <strong>Adobe Experience Cloud</strong> integration</li><li>Robust <strong>offline support</strong> and <strong>multi-channel capture</strong></li><li>Conversational form experiences for social and messaging app engagement</li><li>Strong security and compliance posture</li><li>Modern cloud-native deployment model available</li></ul><h3>Conclusion</h3><p><strong>AEM Forms</strong> stands apart as a powerful, enterprise-grade platform for organizations that require sophisticated, scalable, and secure form experiences. From adaptive design to workflow automation and backend integration, it’s built to support end-to-end digital transformation initiatives.</p><p>If your organization is looking to modernize customer onboarding, service applications, or internal data collection processes, AEM Forms should be at the top of your evaluation list.</p><h3>Ready to Get Started?</h3><p>Learn more about how AEM Forms can transform your digital form strategy. <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/forms-overview/home">Visit Adobe’s official AEM Forms page</a> or reach out to a solution consultant.</p><p>Let me know if you have any question about the content or AEM Forms. Happy to connect!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b63b684923bc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[From Manual Nudges to Automated Intelligence: Driving over 100% Increase in Forum Answer Rates]]></title>
            <link>https://medium.com/@khushwantsehgal/from-manual-nudges-to-automated-intelligence-driving-over-100-increase-in-forum-answer-rates-5f544a75a62d?source=rss-aaca1df59fb8------2</link>
            <guid isPermaLink="false">https://medium.com/p/5f544a75a62d</guid>
            <category><![CDATA[community-management]]></category>
            <category><![CDATA[behavioral-economics]]></category>
            <category><![CDATA[nudge]]></category>
            <category><![CDATA[choice-architecture]]></category>
            <category><![CDATA[customer-support]]></category>
            <dc:creator><![CDATA[Khushwant Sehgal]]></dc:creator>
            <pubDate>Fri, 23 May 2025 08:59:10 GMT</pubDate>
            <atom:updated>2025-05-27T08:13:30.627Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XiCGALSGT0BqABpYAuZ3og.jpeg" /></figure><p>A few quarters ago, during a product and engineering management team discussion on improving customer satisfaction, Director of Engineering voiced a concern that resonated with all of us: “Have you guys noticed how many questions go unanswered in the community forums? The report is looking pretty bad.”</p><p>He was right. We’d all seen the growing backlog of customer questions. Despite having a talented support and development teams, queries often got buried under newer posts. This wasn’t just an internal metric; it meant frustrated customers feeling unheard and valuable opportunities to help (and learn) slipping through our fingers.</p><p>To understand why this was happening, we interviewed numerous developers, product managers, and support consultants. A clear pattern emerged: this challenge wasn’t unique to our Community Forums. It’s a common hurdle for most enterprise forums. The core reason? Product teams — developers, product managers, and support consultants alike — primarily collaborate internally using tools like Slack, Teams, and Email. Proactively monitoring community forums simply wasn’t an ingrained part of their daily workflow.</p><p>The question then became: how could we bridge this gap?</p><h3>Starting Simple: The Manual Nudge Approach</h3><p>Rather than jumping straight to a technical solution, we decided to test a behavioral theory first. Could we influence team behavior through simple, well-placed nudges?</p><p>So we started with the most basic approach possible —We designated one team member (myself) to manually compile lists of unanswered questions from the Experience League forums. Two or three times a week, I’d post these lists to various AEM Forms Slack channels, tagging the corresponding managers with a simple message: “Can you help answer the following customer queries?”</p><figure><img alt="Manual Process" src="https://cdn-images-1.medium.com/max/1024/1*B94qzMTKW8KlrSW4jP-2kA.png" /><figcaption>Manual Process</figcaption></figure><p>This plain-text approach was applying <a href="https://en.wikipedia.org/wiki/Choice_architecture"><strong>choice architecture</strong></a><strong> </strong>in its simplest form. By making the unanswered questions visible, organizing them by date, and framing them as an open request for assistance, we created an environment where team members felt responsible for addressing customer needs.</p><p>The results surprised us. Within weeks, the resolution rate jumped by nearly 35%. Team members started claiming questions in thread replies, and a healthy sense of collaborative responsibility emerged. People began recognizing who had expertise in which areas, and spontaneous mentoring relationships formed.</p><blockquote>“Small and apparently insignificant details can have major impacts on people’s behavior. A good rule of thumb is to assume that “everything matters.”<br>― <a href="https://en.wikipedia.org/wiki/Richard_Thaler"><strong>Richard H. Thaler,</strong></a><strong> </strong><a href="https://hls.harvard.edu/faculty/cass-r-sunstein/"><strong>Cass R. Sunstein</strong></a></blockquote><h3>The Challenges of Manual Curation</h3><p>Despite this success, the manual process had obvious limitations. I was spending significant time each week combing through forums, copying information, and formatting messages. Manually figuring out the best person to tag for each specific question was also time-consuming and imperfect. When I was unavailable, the process would break down. Some weeks, when product priorities shifted, the forum check would get postponed or forgotten.</p><p>As one team member put it, “We love getting these question compilations, but we need to make this sustainable.”</p><h3>Designing for Sustained Engagement</h3><p>When we decided to automate this process, I was careful not to lose the behavioral elements that made the manual process successful. I didn’t just want a script that read forums — I needed to preserve the nudge architecture that was driving engagement. The key was using <strong>LLM models</strong> to analyze question content and ensure the right notification went to the right expert, making the nudge far more effective than a generic alert.</p><p>With access to Jenkins and LLM models from AEM Forms engineering team, I designed a Python-based solution, which we internally dubbed ‘<a href="https://theforumscout.vercel.app/">Forum Scout</a>’, that would:</p><ul><li>Maintain the same direct language in notifications that drove accountability</li><li>Preserve the human-readable formatting that made questions easy to scan</li><li>Include enough context for team members to assess their ability to help</li><li>Deliver insights consistently to maintain the rhythm we had established</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EcTDbrZaUsN_MMr34tKptA.png" /><figcaption>Forum Scout Workflow</figcaption></figure><p>The <a href="https://en.wikipedia.org/wiki/Choice_architecture"><strong>choice architecture</strong></a><strong> </strong>principles remained central: make it visible, make it actionable, and make it targeted to the right people.</p><blockquote>What’s truly compelling about the ‘Forum Scout’ initiative is its sophisticated blend of AI with a deep understanding of human behavior. By thoughtfully applying principles of choice architecture, we didn’t just build a tool; we designed an environment that naturally guides our experts to engage and resolve customer queries. It’s a prime example of how human-centered design can make technology far more effective. — <a href="https://www.linkedin.com/in/saliltaneja/">Salil Taneja, Director — Engineering, Digital Experience, Adobe</a>.</blockquote><h3>Beyond Automation: Creating New Possibilities</h3><p>Once automated, we discovered opportunities that weren’t possible with the manual process. We could now route different question types to specialized channels. We could track resolution rates over time. We could identify trends in customer questions that signaled documentation gaps.</p><p>One unexpected benefit emerged when AEM Forms documentation team started using the historical data to rewrite sections of our self-help content proactively. “You&#39;re not just answering these questions faster,” one of our senior engineering leads mentioned, “You’re preventing some questions from being asked in the first place.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oJz4UxyO1jjP2RgslWzDBQ.png" /></figure><h3>The Results: Numbers Tell the Story</h3><p>Three months into this initiative, both manual and automated, the numbers speak for themselves:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Czb5osE1DUYe_taGC8Ds4g.png" /></figure><p>Forum questions now receive initial responses within an average of 4.2 hours, down from 26 hours previously. The resolution rate has climbed from 46% to 97%.</p><p>A senior consultant recently told me, “I’ve never seen a product team so responsive on forums. It makes me confident in standardizing on your platform.”</p><blockquote>‘Forum Scout’ project brilliantly demonstrates the essence of Nudge theory: by thoughtfully redesigning digital environments, rather than trying to change people, we can drive significant behavioral shifts, achieving remarkable engagement and real-world impact. — <a href="https://www.linkedin.com/in/ganuthula/">Venkat Ram Reddy, Professor of Digital Behavior Dynamics, IIT Jodhpur</a>.</blockquote><h3><strong>Lessons Learned</strong></h3><p>What started as a simple experiment evolved into a powerful tool for customer intelligence and internal efficiency. Our journey offered valuable takeaways for different teams:</p><p><strong>For Product Management Teams:</strong></p><ul><li>Forums are vital listening posts, treat them as essential product components for unfiltered feedback, real-world usage insights, and rapid sentiment tracking.</li><li>Invest in forum health allocate resources to processes (like nudges) and platform features that make forums a demonstrably valuable part of the customer journey.</li><li>Tie forum engagement to business metrics, faster, effective responses boost CSAT, time-to-value, brand perception, and overall product engagement.</li><li>Drive user enablement &amp; adoption via active forums, empower users to self-serve, learn from peers, and overcome hurdles, deepening their engagement.</li></ul><p><strong>For Teams Building AI &amp; Automation Tools:</strong></p><ul><li>Before automating, deeply understand the target behavior or bottleneck by manually validating the problem.</li><li>Make automated notifications and dashboards easy to scan, actionable, and reliable. Consistency builds the trust essential for influencing behavior.</li><li>Set the project to track key indicators (e.g., response/answer rates, support ticket volume) from day 1 to clearly demonstrate your tool’s impact and value.</li><li>Break down challenges into manageable steps (e.g., Identify -&gt; Analyze -&gt; Route -&gt; Notify). Focus on an MVP for one core step first.</li></ul><p>Sometimes, the most impactful tools aren’t new customer features but systems that empower your team to better serve those customers. ‘Forum Scout,’ born from manual effort and behavioral insights, is now invaluable for staying connected to customer needs.</p><figure><img alt="https://threadwise-insights-hub-main.vercel.app/" src="https://cdn-images-1.medium.com/max/1024/1*pIUSURhUt0LkD5kmtLdRfQ.png" /></figure><blockquote>How is your team tackling forum engagement? Have you experimented with similar nudges? Want to discuss this further, adapt these ideas for your own work, or just exchange notes? I’d be <a href="https://www.linkedin.com/in/khushwant-singh-sehgal/">happy to connect</a>. Let us know!</blockquote><blockquote>This blog and the accompanying project code were developed in collaboration with <a href="https://www.linkedin.com/in/shivani-batra-69570b93/"><strong>Shivani Batra</strong></a>. You may reach out to either of us for further questions or discussion.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5f544a75a62d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A Product Manager’s Guide to Early Adopter Programs (key to better products)]]></title>
            <link>https://medium.com/@khushwantsehgal/a-product-managers-guide-to-early-adopter-programs-key-to-better-products-b6e9667daba7?source=rss-aaca1df59fb8------2</link>
            <guid isPermaLink="false">https://medium.com/p/b6e9667daba7</guid>
            <category><![CDATA[management]]></category>
            <category><![CDATA[documentation]]></category>
            <category><![CDATA[feedback]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[product-development]]></category>
            <dc:creator><![CDATA[Khushwant Sehgal]]></dc:creator>
            <pubDate>Thu, 22 May 2025 08:19:50 GMT</pubDate>
            <atom:updated>2025-05-22T08:19:50.901Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*sucaWjk2r_77oL_qxjXe4g.jpeg" /></figure><p>Product managers constantly strive to deliver innovative solutions that resonate with our users. But let’s face it, launching a brand new feature or product can be nerve-wracking. What if only some users like it? What if there are hidden bugs lurking beneath the surface?</p><p>Enter the Early Adopter Program (EAP) — your secret weapon for a smooth and successful launch. Think of it as a real-world testing ground where a select group of enthusiastic customers get a sneak peek at your latest creation. In return for their valuable feedback, they gain early access and a chance to influence the product’s direction.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*v7eHDlMLDMgyM5QQvqqb8Q.png" /><figcaption>Why Early Adopter Program (EAP) Rocks?</figcaption></figure><h3><strong>Why Early Adopter Program (EAP) Rocks?</strong></h3><p>Here’s why you, as a product manager, should be a huge proponent of EAPs:</p><p><strong>Uncover the Bugs Before They Bite:</strong> EAPs act as a bug-squashing boot camp. You get real users testing your features, revealing issues you might have missed in internal testing.</p><p><strong>Feedback is Gold:</strong> EAP participants are your guinea pigs, but the good kind! Their honest feedback provides invaluable insights into user behavior, usability, and feature adoption. This goldmine of information helps you iterate and refine your product before unleashing it on the masses.</p><p><strong>Validation Nation:</strong> EAPs can be a powerful tool for market validation. Seeing how users interact with your new feature can confirm its value proposition and identify potential areas for improvement.</p><p><strong>Customer Champions:</strong> Early adopters are more likely to become brand advocates. They’ve been part of the journey, and their positive experiences can fuel word-of-mouth marketing and build excitement for the official launch.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Fqr24ly3-orjwcp5qryA2A.jpeg" /></figure><h3><strong>Building a Winning EAP</strong></h3><p>Now that you’re convinced (hopefully!), let’s talk about building a stellar EAP. Here’s a roadmap to guide you:</p><ol><li><strong>Know Your Why:</strong> Before diving in, define your EAP’s goals. Are you hunting for bugs, gauging user interest, or validating a specific feature?</li><li><strong>What Does “Success” Look Like?</strong> How will you know if your EAP worked? Are you able to attract and onboard customers to try features under early access program (EAP). For example, onboarding 10 enterprise customers each quarter is a great success number. Maybe success is ‘we found and fixed 5 major usability issues’ or ‘80% of our early adopters said they’d use this daily’.”</li><li><strong>Create EAP documentation page:</strong> List all the EAP innovations on the page, keep the page updated with new innovations, get rid of anything that is not part of early adopter now. Link EAP documentation from multiple locations. Provide customers an email address, or a form to join your EAP program. Make it easy to join the program, &amp; know what’s what. For example, see <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/forms-overview/early-access-ea-features">AEM Forms EAP page</a>.</li><li><strong>Assemble Your Dream Team:</strong> Collaboration is key. Involve engineers to ensure a stable build and the documentation team to create clear instructions for early adopters. Documentation is your first line of defense and engagement driver. PM to drive discussions and recognize patterns. Engineers keep things running smoothly and fix stuff fast. Docs make sure everyone knows how to use it. And you, the PM, to connect all the dots!</li><li><strong>Cast the Net:</strong> Recruit the perfect participants. Look for customers who are engaged, tech-savvy, and comfortable providing constructive feedback. Create a pitch deck. It would making pitching easier and ensure all the product managers are on same page. Your pitch deck is just a simple way to tell them: what they’ll test, what’s in it for them, and what you need from their end.</li><li><strong>Onboard Like a Boss:</strong> Provide clear instructions, training materials, and dedicated support channels for a smooth onboarding experience. A common Slack, Discord, or Microsoft Teams channel is most preferred way to get support and create interaction point between customers and product team these days.</li><li><strong>Feedback Loop Frenzy:</strong> Actively collect feedback through interviews, and dedicated forums. Try find the common thread between multiple customer. When you find those common threads, that’s your cue to build something lots of people will love, not just one person!</li><li><strong>Iterate &amp; Refine:</strong> Don’t just collect feedback, use it! Analyze the data, identify trends, and incorporate the learnings to improve your product.</li><li><strong>Defined deadline: </strong>Run the program for each customer with tight deadlines. For example, if you have enterprise software, for a single capability, engagement with a customer should end within 6–8 weeks</li></ol><h3><strong>What Could Go Wrong? (And How to Fix It)</strong></h3><ul><li><strong>People get tired?</strong> Keep engagement short and sweet. Don’t ask for too much meetings.</li><li><strong>Secret stuff gets out?</strong> Ask them nicely to keep it hush-hush (sometimes people use simple agreements for this).</li><li><strong>Users expect everything to be perfect?</strong> Remind them it’s an early look. Work with your engineers to make software as stable as possible.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fSZxw2F-xr61YIQWlgFCGA.jpeg" /></figure><h3><strong>Spreading the Word</strong></h3><p>Let the world (or at least your target audience) know about your awesome EAP! Leverage your marketing channels — blog posts, social media, targeted emails, internal slack channels, organic SEO — to reach potential participants. Create a dedicated landing page outlining program benefits, the application process, and available features. Use top documentation pages, forums, and post on your company blog or send targeted emails to spread word about the program.</p><h3><strong>The EAP: A Win-Win Proposition</strong></h3><p>EAPs benefit both you and your customers. You get valuable insights and a more polished product. Your customers get early access, shape the future of the software they love, and potentially score some exclusive perks.</p><p>So, what are you waiting for? Embrace the power of the EAP and launch your next product with confidence!</p><p>AEM Forms EAP program went live in Jan 2024. Since than it has received 250+ requests to access various innovations being delivered via the program.</p><h3><strong>How to Join the AEM Forms EAP</strong></h3><p>We’re looking for enthusiastic AEM Forms <a href="https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/forms/forms-overview/early-adopter-ea-features">users who are eager to test new features and provide constructive feedback</a>. If you’re interested in joining the program, simply send an email expressing your interest to aem-forms-ea@adobe.com.</p><p>Feel free to reach me at <a href="https://twitter.com/khushwantsehgal">@khushwantsehgal</a> with your queries, if any.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b6e9667daba7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Shall Save for Myself and Spend for the Economy — Kids Edition]]></title>
            <link>https://medium.com/@khushwantsehgal/i-shall-save-for-myself-and-spend-for-the-economy-kids-edition-600acc83b9e0?source=rss-aaca1df59fb8------2</link>
            <guid isPermaLink="false">https://medium.com/p/600acc83b9e0</guid>
            <category><![CDATA[spending]]></category>
            <category><![CDATA[saving]]></category>
            <category><![CDATA[children]]></category>
            <category><![CDATA[economics]]></category>
            <dc:creator><![CDATA[Khushwant Sehgal]]></dc:creator>
            <pubDate>Sun, 18 May 2025 11:22:11 GMT</pubDate>
            <atom:updated>2025-05-20T09:24:46.677Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eeaB1dsDRY6pHpznX9ETuA.jpeg" /></figure><h3>I Shall Save for Myself and Spend for the Economy — Kids Edition</h3><p>Imagine you have a magical piggy bank and a super fun toy shop.<br>Every time you get some money, like when you help with chores or get a gift, you have a few choices:</p><ol><li>Put some in your <strong>magical piggy bank</strong> to save for yourself.</li><li>Put some in a <strong>Super Saver Town Bank</strong> where it can grow and help others.</li><li>Spend some at the <strong>super fun toy shop</strong> to help your town.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nFM-mFfsftBlk2n9I3CwFA.jpeg" /></figure><h3><strong>Why Save for Yourself?</strong></h3><p>When you put your coins in the magical piggy bank, it’s like creating your own secret treasure chest! You’re saving up for something special you really want — like a bicycle or a cool robot. When you finally open it, all that money is there for <em>you</em>.</p><p>Saving this way also helps you feel safe. If you ever need money for something important, you can always rely on your savings. It’s like having a superhero shield for surprises!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cepoCHgY9Jh_G-Tx0qs3XQ.jpeg" /></figure><h3>Why Save (and Invest!) in a Super Saver Town Bank?</h3><p>Now, imagine your piggy bank savings get bigger, or you want your money to do even <em>more</em> magic. That’s where the <strong>Super Saver Town Bank</strong> comes in!</p><p>When you put your money in a Town Bank, it doesn’t just sit there like treasure hidden under your bed. The bank keeps it safe for you, and here’s the extra magic: they let other people in town, like grown-ups who want to start a new bakery or build a bigger toy factory, <em>borrow</em> some of that money.</p><ul><li><strong>Helping Others:</strong> This borrowing helps them create new things and jobs!</li><li><strong>Making Your Money Grow:</strong> Because the bank helped them, when they pay the money back, they often pay a little extra. Some of that extra can make <em>your</em> savings in the bank grow a little bit too, without you doing anything! This is called <strong>investment</strong> — it’s like planting a money seed that grows into a slightly bigger plant.</li></ul><p>So, money in a Town Bank is working hard! It’s helping the town and potentially growing your savings at the same time. It’s not “sleeping money”; it’s “busy money”!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Y4M_Gwd84GevhcPOrSEecg.jpeg" /></figure><h3><strong>Why Spend for the Economy?</strong></h3><p>Now, spending is awesome too! When you buy a new toy, you make the toy shop happy, and they use the money to bring even cooler toys next time. They also pay the workers who help make those toys. Those workers then go to the grocery store, the grocery store buys more food, and everyone gets what they need.</p><p>When you spend, your money travels all around the town, helping people do their jobs and bring more fun things to life. It’s like sending your coins on a happy adventure!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iIl4TZAXaXvdtRKsY1-phg.jpeg" /></figure><h3><strong>Be Hero of your City</strong></h3><p>Here’s the coolest part: when you save (in your piggy bank or the Town Bank!) and spend for the economy, you are actually a hero for your town!</p><ul><li>When you save in your <strong>piggy bank</strong>, you’re learning a super skill and planning for your dreams!</li><li>When you save and invest in the <strong>Super Saver Town Bank</strong>, your money joins up with other people’s savings. This big pool of “busy money” helps leaders and businesses build amazing things like playgrounds, schools, libraries, and new shops. It helps them get the resources to make your town more fun, safe, and full of opportunities. This is how savings help the whole town grow and invest in its future!</li><li>When you <strong>spend</strong>, you help people have jobs, make new toys, and fill up the shops with exciting things. Your money brings smiles to workers, shop owners, and families all around!</li></ul><p>Saving in your piggy bank is great for your goals. Saving and investing through a Town Bank helps your money work for the town’s future (and maybe grow a bit for you!). Spending makes your town glow with activity right now.</p><p>That’s why when you get your next shiny coin, you can think, ”I shall save for myself, invest for our future, and spend for the economy!” because you know you’re doing something truly magical.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=600acc83b9e0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A product management director shared with me, how to structure first call with enterprise customers?]]></title>
            <link>https://medium.com/@khushwantsehgal/what-it-takes-your-first-discovery-call-77dda5aeaca3?source=rss-aaca1df59fb8------2</link>
            <guid isPermaLink="false">https://medium.com/p/77dda5aeaca3</guid>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[product-manager]]></category>
            <category><![CDATA[product-manager-interview]]></category>
            <category><![CDATA[discovery-call]]></category>
            <category><![CDATA[product-management-skills]]></category>
            <dc:creator><![CDATA[Khushwant Sehgal]]></dc:creator>
            <pubDate>Tue, 22 Oct 2024 04:42:31 GMT</pubDate>
            <atom:updated>2026-02-14T15:05:11.619Z</atom:updated>
            <content:encoded><![CDATA[<iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FF5ZgcA2JGqU%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DF5ZgcA2JGqU&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FF5ZgcA2JGqU%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/d4ac2e00d3bf0fb81d9c6b1a0073e473/href">https://medium.com/media/d4ac2e00d3bf0fb81d9c6b1a0073e473/href</a></iframe><p>Stepping into your new role as a product manager feels like embarking on an epic adventure. Fresh from your MBA program, you are armed with knowledge, eager to prove yourself, and ready to tackle challenges head-on.</p><p>Today, you face your first major quest: the discovery call. You are tasked to interview customer to find the problems and their usecases.</p><p>The discovery call isn’t just a conversation; it’s a journey to uncover your customer’s true needs, pain points, and dreams. Let’s explore how to invest in your customer’s emotional, personal, and business success during this critical call.</p><h3>Emotional Success</h3><p>Connect on a human level. Recognize that behind every professional role there’s an individual with emotions, motivations, and aspirations. Ask open-ended questions to understand their emotional drivers:</p><ul><li>What is your title and role? Probe to understand nitty-gritties of the role. Show interest in learning about your customer.</li><li>What department or area is the customer responsible for?</li><li>What are they trying to achieve?</li><li>Who do they report to?</li><li>What does the organizational hierarchy look like?</li><li>What problem are they facing?</li><li>How frustrated are they with the problem?</li><li>Which is the biggest problem? Make them live the journey.</li></ul><p>Establish trust. Show empathy and genuinely listen to their concerns. Repeat back what they’ve shared to confirm your understanding. This validates their feelings and demonstrates that you are invested in their emotional success.</p><p>Make the discovery call a positive experience. Be enthusiastic and supportive. Acknowledge their efforts and achievements. Assure them that your goal is to make their professional life smoother and more rewarding.</p><h3>Personal Success</h3><p>Beyond their professional responsibilities, your customers have personal goals and ambitions. Ask questions like:</p><ul><li>What are their KPIs?</li><li>What will make the customer look like a hero?</li><li>What does success look like for your customer?</li><li>Think how can your product make your customer a hero in his/her boss’s eyes? Would reduces cost help? or improved velocity or better conversion rate would bring more acolades? Become invested in personal success of your customer.</li></ul><p>Show that you’re invested in their personal growth. Offer resources, training, or insights that can help them develop professionally. This strengthens your relationship and positions your product as a valuable tool for their career advancement.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QM06RJRepERhRCIqAfFk5A.png" /></figure><h3>Business Success</h3><p>Your customer’s success is tied to their organization’s success. Begin by exploring their business objectives:</p><ul><li>What are your customer’s top business priorities?</li><li>How does your customer’s team measure success?</li><li>What does your customer use your product for? To make themselves better? Compete better? Socialize it? Find out.</li><li>How many steps do they follow to solve the problem today?</li><li>What is their current workflow?</li><li>How often do they follow their current workflow?</li><li>Debug from point A to point B with them.</li><li>Go deeper into their current challenges:</li><li>Ask about their workflow, the steps they take to achieve their goals, and the obstacles they face. This is about identifying opportunities for your product to make a meaningful impact. Make them live the journey. Ask for their experience. How frequently are they facing the problem? Dig deeper.</li></ul><p>Your first discovery call is a critical step in your journey as a product manager. By investing in your customer’s emotional, personal, and business success, you build strong, lasting relationships and set the stage for mutual success. Embrace the adventure, learn from each interaction, and let your curiosity guide you. Welcome to the exciting world of product management — your quest has just begun!</p><p>Feel free to reach me at <a href="https://twitter.com/khushwantsehgal">@khushwantsehgal</a> with your suggetions and queries.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=77dda5aeaca3" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>