<?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 Pamornpol Jinachitra on Medium]]></title>
        <description><![CDATA[Stories by Pamornpol Jinachitra on Medium]]></description>
        <link>https://medium.com/@pamornpol?source=rss-1c166df458de------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*kIWdVACa6gXHmBZL.jpg</url>
            <title>Stories by Pamornpol Jinachitra on Medium</title>
            <link>https://medium.com/@pamornpol?source=rss-1c166df458de------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 07:10:30 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@pamornpol/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[Why KPI Misalignment Is Quietly Killing Your AI Adoption]]></title>
            <link>https://pamornpol.medium.com/why-kpi-misalignment-is-quietly-killing-your-ai-adoption-f0249df1bb37?source=rss-1c166df458de------2</link>
            <guid isPermaLink="false">https://medium.com/p/f0249df1bb37</guid>
            <category><![CDATA[ai-business-strategy]]></category>
            <category><![CDATA[ai-for-business]]></category>
            <category><![CDATA[ai-business-integration]]></category>
            <category><![CDATA[practical-ai]]></category>
            <dc:creator><![CDATA[Pamornpol Jinachitra]]></dc:creator>
            <pubDate>Mon, 11 May 2026 02:43:11 GMT</pubDate>
            <atom:updated>2026-05-11T02:43:11.547Z</atom:updated>
            <content:encoded><![CDATA[<p>Practical AI#4</p><h4><strong>The technology isn’t the problem. The incentive system is.</strong></h4><p>One of the most consistent failure patterns I’ve seen across AI implementations — in financial services, insurance, and beyond — isn’t a model accuracy problem or a data quality problem.</p><p>It’s a KPI problem.</p><p>After six years of deploying AI systems inside large organizations, I’ve come to believe that how you measure people matters more than what technology you deploy. Get the incentives wrong, and even the best AI system will sit unused. Get them wrong in the other direction, and you’ll burn through your annual AI budget in four months.</p><p>Both of these things have actually happened. Let me walk through both.</p><h3>When Existing KPIs Create Resistance</h3><p>Picture a loan officer at a bank. Before AI, her job was measured on cases processed per day, time-per-case, and error rate — with a zero-error-tolerance policy on top of that.</p><p>Now her company rolls out an AI-assisted document review tool. Management is excited. The vendor demo looked great. But six months in, adoption is nearly zero.</p><p>Why?</p><p>When I talked to frontline staff in situations like this, the reasoning was completely rational:</p><p><strong>“If AI makes a mistake, who gets blamed?”</strong> The error still shows up on her record. The KPI doesn’t care whether a human or a machine caused it.</p><p><strong>“I don’t have time to experiment.”</strong> Her SLA is already stressful enough. Adding an unfamiliar tool mid-workflow isn’t a learning opportunity — it’s a risk to her performance score.</p><p><strong>“Why would I help make it smarter?”</strong> Providing feedback to improve the AI takes time. And the better the AI gets, the more it threatens her job. The incentive to contribute is essentially negative.</p><p><strong>“I’ll only use it when it’s perfect.”</strong> Staff often set impossibly high accuracy bars before they’ll adopt — not out of stubbornness, but because their KPIs leave no room for the learning curve that every AI system requires.</p><p>“<strong>It will help you work faster.</strong>” Even with this widely used statement of encouragement, the follow-up question was whether they would have to do more cases then? The time saved has not been articulated clearly how it will benefit them.</p><p>None of this is irrational behavior. These are intelligent people responding logically to the incentive structure around them. The system is working exactly as designed — just not in the direction anyone intended.</p><p>No matter how good the technology gets, if the incentive structure signals <em>“don’t take risks,”</em> adoption will stall.</p><h3>When AI-Friendly KPIs Go Too Far</h3><p>The opposite failure is less common, but increasingly worth watching.</p><p>Earlier this year, Uber’s CTO revealed that the company burned through its entire 2026 AI budget in just four months. The trigger: after announcing a push for engineers to maximize AI coding tool usage — with KPIs tied to adoption — roughly 95% of engineers shifted their workflows to tools like Claude Code and Cursor almost overnight.</p><p>Uber’s case may not represent true misalignment in the classical sense. The productivity gains might well justify the cost. But the episode illustrates something important: <strong>when you measure AI usage volume without measuring what that usage produces, you’ve created a system that can be gamed — even unintentionally.</strong></p><p>There are also reports of Meta implementing token usage dashboards as internal ranking mechanisms. To be fair, Meta likely has strategic reasons beyond simple adoption metrics — including data collection for model training and creating strong behavioral nudges. The leaders there aren’t naive about misalignment. But for most organizations without Meta’s specific context, measuring AI engagement by token consumption is a warning sign, not a model to copy.</p><p>Volume is not value.</p><h3>What Can We Do?</h3><p>Based on what I’ve seen across many industries and organizations, here’s where to focus:</p><p><strong>Create a proper sandbox — with no zero-error-tolerance policy inside it.</strong></p><p>Before asking staff to use AI in live workflows, give them a safe space to experiment. This can be as simple as a shadow-run environment that mirrors production without affecting real outcomes, or a live workflow with limited scope but acceptable risk and tolerance. The goal is to build familiarity, gather feedback, and let people develop genuine intuition about where AI helps and where it doesn’t — without their performance scores on the line.</p><p><strong>Align incentives with outcomes, not with usage.</strong></p><p>Don’t measure whether someone used AI. Measure whether their output improved — in volume, quality, speed, or business value. If an underwriter processes more claims with the same accuracy, that’s the signal you want. How they got there is secondary.</p><p><strong>Reward people who catch AI mistakes.</strong></p><p>This one is underutilized. Gamifying error-catching — recognizing or rewarding the staff member who identifies the most AI errors and submits useful corrections — transforms frontline employees from reluctant adopters into active contributors to model improvement. It reframes the human role from “being replaced” to “being the quality layer.”</p><p><strong>Give business unit heads skin in the game.</strong></p><p>KPIs for AI adoption shouldn’t just sit with individual contributors. BU leaders need to own them. When a team head is accountable for their unit’s AI adoption outcomes — not just their traditional performance metrics — the dynamic shifts completely. It stops being an IT initiative and becomes a business priority.</p><p><strong>Let the savings flow to the people generating them.</strong></p><p>If AI helps a team handle 30% more cases, and some portion of that efficiency translates into additional revenue or cost savings — consider sharing a slice of that with the team. Even a modest performance bonus tied to AI-enabled productivity can shift the emotional framing from <em>“this tool threatens my job”</em> to <em>“this tool earns me more.”</em></p><h3>The Underlying Principle</h3><p>AI investment requires capital. But capital alone doesn’t create adoption.</p><p>The organizations I’ve seen successfully integrate AI at scale aren’t the ones with the most advanced models or the biggest budgets. They’re the ones that took the time to ask: <em>“What are we actually rewarding people for? And does that align with what we want them to do?”</em></p><p>If your KPIs haven’t been updated since before your AI initiative launched, they’re probably working against you.</p><p>Fix the incentive layer first. The technology will likely follow.</p><p><em>This is part of my ongoing series, Practical AI for Business — practitioner perspectives on enterprise AI adoption from the ground up.</em></p><p><em>If this resonated, follow along for more: #PracticalAI #AIforBusiness</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f0249df1bb37" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Start Adopting AI When You Don’t Know Where to Begin]]></title>
            <link>https://pamornpol.medium.com/how-to-start-adopting-ai-when-you-dont-know-where-to-begin-802486dadc8c?source=rss-1c166df458de------2</link>
            <guid isPermaLink="false">https://medium.com/p/802486dadc8c</guid>
            <category><![CDATA[practical-ai]]></category>
            <category><![CDATA[gen-ai-for-business]]></category>
            <category><![CDATA[ai-for-business]]></category>
            <dc:creator><![CDATA[Pamornpol Jinachitra]]></dc:creator>
            <pubDate>Mon, 04 May 2026 13:39:48 GMT</pubDate>
            <atom:updated>2026-05-04T13:39:48.890Z</atom:updated>
            <content:encoded><![CDATA[<h4>Practical AI#3</h4><h4>Most organizations don’t have a technology problem. They have a clarity problem.</h4><p>One of the most common questions I get from executives isn’t about which AI vendor to choose or how much to budget. It’s more fundamental than that:</p><p><em>“We know we need to do something with AI. We just don’t know where to start.”</em></p><p>What makes this question harder than it sounds is that there often isn’t a ready-made answer to copy. In many markets — particularly across Southeast Asia — genuine, detailed case studies of enterprise AI adoption are still scarce. Even the major global consulting firms, when you push past the polished frameworks, are often offering repackaged common sense rather than hard-won operational insight.</p><p>So here’s the starting framework I give executives when they’re standing at the beginning with no map: <strong>Know the technology. Know your organization. Then bring them together.</strong></p><h3>Know the Technology — Before You Trust It</h3><p>Most executives I work with have seen impressive AI demos. A few have been genuinely wowed. Some have been burned.</p><p>The problem with demos is that they’re optimized for the best-case scenario — clean inputs, predictable queries, ideal conditions. Real organizational data is messier. Real use cases are more ambiguous. And real AI systems, even the best ones available today, hallucinate, misinterpret context, and fail in ways that don’t show up in a vendor presentation.</p><p>The fastest way to close this gap is to simply use the tools yourself.</p><p>Not in a structured evaluation. Not through a procurement committee. Just open ChatGPT, Gemini, or Claude and start pushing on them. Try to break them. Feed them the kind of documents and questions your organization actually deals with. Notice where they’re surprisingly capable and where they fall apart unexpectedly. Learn from a lot of readings that do not only discuss successful use cases, absorb information about limitations and some failed cases too.</p><p>This matters because executives who’ve never personally used these tools tend to miscalibrate in both directions — overestimating AI’s reliability in high-stakes decisions, and underestimating how much operational lift it can provide in repetitive, well-defined tasks.</p><p>Firsthand experience won’t make you an AI expert. But it will make you a much better buyer, a much sharper client, and a much more effective decision-maker when it comes time to commit resources.</p><h3>Know Your Organization — Before You Commit to a Direction</h3><p>Once you have a realistic model of what AI can and can’t do, the second step is turning that lens inward.</p><p>This means doing an honest audit of your own operations — not from a strategy document, but from the ground up. A few questions worth sitting with:</p><p><strong>Where is time actually going?</strong> Not where you think it’s going based on org charts and job titles, but where hours genuinely accumulate on the ground. The answer often surprises leadership.</p><p><strong>Where do errors cluster?</strong> Manual, repetitive, high-volume processes tend to be both the most error-prone and the most amenable to AI-assisted improvement. That overlap is usually where the early ROI lives.</p><p><strong>Is your data actually ready?</strong> AI needs data at the right time, in the right format, with the right access controls. Many organizations discover, only after committing to an AI project, that their data infrastructure isn’t set up to support it. This is worth assessing honestly before you’re mid-implementation.</p><p><strong>What’s the change appetite of your people?</strong> Technology is rarely the binding constraint in AI adoption. Culture and incentive structures usually are. An organization whose frontline staff feel threatened by automation will resist AI in ways that no vendor roadmap accounts for.</p><h3>Why Copying Others Usually Falls Short</h3><p>There’s an understandable temptation to look sideways — to find out what a competitor or an industry peer is doing with AI and replicate it.</p><p>The problem is that organizational context isn’t portable.</p><p>What works for a large retail bank with thousands of standardized loan applications may not translate to a mid-sized life insurer with complex underwriting decisions and long policy cycles. The surface-level use case might look similar. The underlying process, data structure, regulatory environment, and workforce dynamics often aren’t.</p><p>Benchmarking has its place. But it’s a starting point for questions, not a substitute for the harder work of understanding your own situation.</p><h3>The Underlying Logic</h3><p>Know the technology well enough to have realistic expectations. Know your organization well enough to identify where those capabilities genuinely meet your needs.</p><p>That intersection — between what AI can actually do and what your organization actually needs — is where sustainable AI adoption begins.</p><p>Everything else, the vendor selection, the implementation approach, the change management — flows more naturally once that foundation is in place.</p><p><em>Part of the Practical AI for Business series — practitioner perspectives on enterprise AI adoption.</em></p><p><em>#PracticalAI #AIforBusiness</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=802486dadc8c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Accuracy Is Not Everything. Real Value Is.]]></title>
            <link>https://pamornpol.medium.com/accuracy-is-not-everything-real-value-is-fd95a45bef19?source=rss-1c166df458de------2</link>
            <guid isPermaLink="false">https://medium.com/p/fd95a45bef19</guid>
            <category><![CDATA[practical-ai]]></category>
            <category><![CDATA[gen-ai-for-business]]></category>
            <category><![CDATA[ai-for-business]]></category>
            <dc:creator><![CDATA[Pamornpol Jinachitra]]></dc:creator>
            <pubDate>Fri, 10 Apr 2026 08:32:37 GMT</pubDate>
            <atom:updated>2026-04-10T08:32:37.187Z</atom:updated>
            <content:encoded><![CDATA[<h3>Practical AI for Business #2</h3><p><em>There’s one question I hear in almost every enterprise AI pitch. It’s a reasonable question — but the way it gets asked, and answered, rarely leads anywhere useful.</em></p><p><strong>TL;DR</strong> Accuracy is a valid starting point, but it’s an incomplete — and often misleading — proxy for actual AI performance. Fixating on a single number leads to bad vendor comparisons, misaligned expectations, and deployments that look good on paper but fail in practice. What matters is whether the metric reflects your real use case, and whether the deployment is designed to extract genuine value from whatever capability exists.</p><h3>What Are We Actually Measuring?</h3><p>Walk into any enterprise AI evaluation meeting, and someone will ask: <em>“What’s the accuracy of this model?”</em></p><p>It’s not a wrong question. But in practice, the answer almost never means what the person asking intends it to mean.</p><p>Technically, accuracy is an aggregate metric — it combines a model’s ability to correctly identify true positives with its ability to correctly identify true negatives. Straightforward enough. But the number only carries meaning in context: <em>what data was it measured on, and does that data actually reflect how you intend to use the system?</em></p><p>When Vendor A claims 99% accuracy and Vendor B claims 90%, those figures cannot be meaningfully compared unless both were evaluated on the same dataset, at the same difficulty level, against the same task definition. In most pitch meetings, none of those conditions hold. The benchmarks are different, the test sets are curated differently, and neither may bear much resemblance to your actual documents, your actual users, or your actual edge cases.</p><p>A few examples that come up regularly in practice:</p><p><strong>Handwriting recognition.</strong> Vendors frequently advertise 90%+ accuracy on handwritten text. What they’re often measuring is clean, structured prose — essay-style writing with linguistic context that makes ambiguous characters guessable. Ask that same model to read names on a free-form intake form, and 90% becomes an aspiration rather than a baseline.</p><p><strong>Anomaly detection in medical imaging.</strong> Suppose a condition appears in 1 out of every 100 cases. A model that predicts “negative” for every single input will achieve 99% accuracy — technically correct, entirely useless. This is a well-known failure mode, and it’s why precision and recall are the appropriate metrics for imbalanced classification tasks. Aggregate accuracy, in this context, actively obscures the problem.</p><p><strong>Generative AI response quality.</strong> This one is more philosophically awkward. Accuracy implies a ground truth — a correct answer against which outputs are measured. But for most generative tasks, the “correct” answer is partly subjective, and different evaluators will disagree on what counts as right. Unless you constrain the model to multiple-choice responses, traditional accuracy measurement breaks down almost entirely.</p><h3>Accuracy Is One Snapshot. Act Accordingly.</h3><p>Even the most rigorous benchmarks — the standardized evaluations that leading AI labs use to compare models publicly — should be treated as a single data point, not a verdict.</p><p>GPT, Claude, Gemini, and their peers are measured against academic and industry-standard benchmarks, and those results are published and discussed widely. But any given model tends to outperform on some benchmarks and underperform on others. A model that leads on reasoning tasks may lag on multilingual performance. A model that scores well on structured document extraction may struggle with open-ended synthesis.</p><p>This creates a real challenge for organizations deploying AI in specific linguistic or cultural contexts. Thai-language benchmarks, for instance, are sparse. The impressive headline numbers from global AI evaluations are typically generated on English-heavy datasets. That doesn’t mean these models are ineffective in Thai — but it means the published figures provide limited guidance on what to expect. At minimum, any model you’re considering for a non-English context should demonstrate meaningful multilingual capability in its published evaluations. That’s a necessary, if not sufficient, condition.</p><p>Beyond accuracy, there are metrics better suited to evaluating generative AI in enterprise settings:</p><ul><li><strong>Relevance</strong> — Does the output actually address the question being asked, or does it respond to a plausible interpretation of the question that doesn’t match what the user needed?</li><li><strong>Groundedness</strong> — Does the response stay anchored to the source material, or does the model drift into plausible-sounding confabulation?</li></ul><p>Think of it as the blind men and the elephant, each touching different parts of the animal getting different impressions. Each metric reveals a different part of the picture. No single measurement tells you whether the system is actually fit for purpose — but several measurements, taken together, get you meaningfully closer.</p><p>Leading AI organizations are already moving in this direction, shifting toward metrics with direct operational meaning: task completion rates versus industry professionals, workflow automation coverage, tool-calling reliability. These are harder to benchmark cleanly, but they reflect the questions that actually matter to a business.</p><h3>The Number Is Never the Point</h3><p>Here is the harder truth, and the one most organizations are slow to internalize: no matter what accuracy figure a model achieves, it will not deliver value if the deployment is poorly designed.</p><p>A model running at 85% accuracy, embedded in a thoughtfully redesigned workflow with appropriate human review at the right checkpoints, will consistently outperform a 99% model bolted onto a process that wasn’t built to use it well.</p><p>I’ve seen both sides of this play out directly.</p><p><strong>Case one:</strong> An organization came into a project insisting on 95% accuracy as a hard requirement before sign-off. Committing to that figure before seeing the actual documents — with all their variation in quality, format, and legibility — was, frankly, a commitment no one could honestly make. But once the work began, something more useful happened: we built a shared understanding of where the model performed well, where it struggled, and why specific errors occurred. That understanding drove process redesign. By the time the deployment stabilized, measured accuracy was sitting around 85%. And the team had reduced their manual processing workload by nearly 50%. The number looked worse. The outcome was significantly better.</p><p><strong>Case two:</strong> A different organization insisted their AI must achieve 97% accuracy on a predefined test set before it could go live. The threshold wasn’t unreasonable in isolation — but the path taken to reach it was. The model was progressively fine-tuned against that specific test set until it cleared the bar. What emerged was a textbook case of overfitting: the system had learned to perform well on questions it had seen before, at the expense of its ability to handle the novel, varied inputs that real users actually submit. In production, queries outside the training distribution failed badly. The score looked excellent. The deployment did not.</p><h3>The Right Question to Ask</h3><p>The question worth asking in an AI evaluation isn’t simply <em>“what’s the accuracy?”</em></p><p>It’s: <em>“What was this measured on, does that reflect our real use case, and how are we planning to design the deployment around whatever capability exists?”</em></p><p>That reframe shifts the conversation from a single number — which is always a simplification — toward the combination of measurement, context, and process design that actually determines whether an AI system delivers value.</p><p>The accuracy figure tells you something. It just doesn’t tell you enough.</p><p><em>The author is CEO of AIGEN, an AI-as-a-Service company based in Thailand serving the financial and insurance sectors. Practical AI for Business is a series on enterprise AI adoption from the front lines.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fd95a45bef19" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[State of AI for Business, Q1 2026 — and What It Means for the Labor Market]]></title>
            <link>https://pamornpol.medium.com/state-of-ai-for-business-q1-2026-and-what-it-means-for-the-labor-market-a74ab6fc4365?source=rss-1c166df458de------2</link>
            <guid isPermaLink="false">https://medium.com/p/a74ab6fc4365</guid>
            <category><![CDATA[practical-ai]]></category>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[ai-for-business]]></category>
            <dc:creator><![CDATA[Pamornpol Jinachitra]]></dc:creator>
            <pubDate>Tue, 31 Mar 2026 11:49:57 GMT</pubDate>
            <atom:updated>2026-03-31T11:49:57.288Z</atom:updated>
            <content:encoded><![CDATA[<p>Practical AI for Business #1</p><p><em>Three months into 2026, here’s what I’m seeing from the front lines — as someone who tracks AI developments, uses these tools daily, and pitches them to enterprise clients.</em></p><p><strong>TL;DR</strong> Everything emerging since the start of the year points toward AI displacing human labor — not as a distant possibility, but as a near-term reality. The remaining question is how quickly people and organizations adapt: will they be replaced, or merely displaced? More urgently, Agentic AI is rapidly closing the economic value gap that let many organizations rationalize inaction. Those who haven’t adjusted their mindset may already be the proverbial frog in boiling water.</p><h3>The State of Agentic AI in 2026</h3><p>Since January, there has been a remarkable acceleration in AI releases — several of which feel less like incremental updates and more like genuine turning points, both in capability and in the mainstreaming of agentic AI.</p><p><strong>Claude Cowork</strong> demonstrated that AI can now <em>act</em>, not just respond. Beyond answering questions or drafting content, it manages files, connects to external applications, schedules meetings, updates tasks, and can even trigger the creation of Figma designs — all from a single instruction.</p><p><strong>Autonomous agent frameworks</strong> (OpenAI’s direction included) are giving developers the closest thing yet to a Jarvis-like system — an AI that doesn’t just follow instructions step by step, but resolves problems independently, without being taught each sub-task by hand.</p><p><strong>Claude Code</strong> sent shockwaves through SaaS valuations for good reason: straightforward applications are now buildable by almost anyone. Personally, I’m reconsidering a CRM subscription I’ve held for years. The platform essentially does card entry and funnel tracking. I’ve found that a well-structured spreadsheet handles the data just fine — and Claude can build the dashboard and run the funnel analysis on top of it. Better ROI by a wide margin.</p><h3>So Why Aren’t More Businesses Seeing Results from AI?</h3><p>Setting aside organizations that are genuinely early in their exploration — that’s understandable and expected — the more interesting question is why AI adoption remains shallow in companies that have been at it for a while.</p><p>BCG’s research offers a pointed diagnosis: many large organizations aren’t seeing returns because they’ve been doing many <em>piecemeal </em>AI all over the company or in BCG’s own terminology <a href="https://theaiinnovator.com/bcg-leader-most-banks-still-have-a-long-way-to-go-on-ai/#:~:text=%E2%80%9CThere&#39;s%20an%20element%20of%20letting%20a%20thousand,&#39;Are%20you%20doing%20anything%20in%20gen%20AI?">“A thousand flowers bloom”</a>. They insert AI as a supporting tool into isolated workflows rather than rethinking the workflow itself. There’s no scale, no cross-functional context, and no process redesign to match. I’d call this the <strong>+AI</strong> approach — AI added on top of existing operations.</p><p>The approach that actually delivers results is the inverse: <strong>AI+</strong>, or what some call AI-first. The distinction matters enormously.</p><p>The term “human-centered AI” is worth examining carefully here. In the context of ethics and governance — protecting individual rights, preventing discrimination, ensuring AI outcomes don’t harm people — it’s an essential principle. But when organizations apply it to mean “the existing human user’s workflow comes first,” the results tend to be predictable:</p><ul><li>Processes aren’t redesigned, so AI is forced to fit into workflows it wasn’t built for</li><li>Employees face a quiet conflict of interest — faster AI output can mean heavier workloads, not lighter ones</li><li>User requirements become a ceiling rather than a floor: demanding AI perform tasks that humans themselves can’t reliably do</li></ul><p>AI-first thinking removes these constraints. I’ll explore what that looks like in practice in a future piece.</p><p>The trajectory I’ve watched over the past few years looks something like this:</p><blockquote><strong><em>Before AI:</em></strong><em> 10 steps. Each one slow, manual, and error-prone.</em></blockquote><blockquote><strong><em>2023–2024:</em></strong><em> AI handles roughly 5 of those steps — but inconsistently, requiring careful human review at each stage. Economic value exists in theory, but the overhead of supervision keeps ROI modest. Organizations respond with “human-in-the-loop” policies. Results remain underwhelming.</em></blockquote><blockquote><strong><em>Now:</em></strong><em> AI handles all 10 steps end-to-end. It can even review its own output. And while errors still happen, the economic case for removing humans from the process entirely — at least for defined, structured workflows — is no longer deniable.</em></blockquote><h3>What This Means for the Labor Market</h3><p>The framing of AI as “just a tool” persists in many expert circles, and it carries some truth. But what we’re seeing in 2026 stretches that framing well past its useful limits. A tool that thinks, acts, and delivers results without requiring meaningful human input at any stage isn’t a tool in any conventional sense.</p><p>Why, then, haven’t more companies visibly reduced headcount? From what I’ve observed, the explanations fall into a few buckets:</p><ul><li>Some already have, quietly — it simply isn’t making headlines</li><li>Some are seeing genuine productivity gains but are choosing to redeploy people rather than cut them, at least for now</li><li>And some genuinely haven’t gotten far enough in AI adoption for it to matter yet</li></ul><p>What’s shifting in 2026 is that the first category is becoming harder to ignore. Block recently announced a 40% reduction in workforce this year, explicitly citing internal intelligence tools as a contributing factor. Earlier waves of tech layoffs — Amazon and others — were easier to attribute to Covid-era over-hiring. That explanation has increasingly less cover. AI’s role in these decisions is becoming undeniable, and the trend will reach markets like Thailand within five years. Possibly sooner.</p><p>The argument that upskilling can outpace displacement is still sound in principle. Humans are, historically, a remarkably adaptive species. We absorbed the Industrial Revolution. We absorbed the internet. New categories of work emerged each time.</p><p>But I find myself sitting with a harder question: the people most exposed to this shift — those in administrative roles, data entry, document processing, routine knowledge work — will they have the runway to adapt? Each prior disruption played out over decades. This one is moving faster than any before it.</p><p>That, more than any benchmark or product announcement, is what I keep coming back to.</p><p><em>The author is CEO of AIGEN, an AI-as-a-Service company based in Thailand serving the financial and insurance sectors. This series, Practical AI for Business, shares observations from the front lines of enterprise AI adoption.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a74ab6fc4365" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[According To Me, The Series “Measure What Matters” by John Doerr]]></title>
            <link>https://pamornpol.medium.com/according-to-me-the-series-measure-what-matters-by-john-doerr-81c1040d488d?source=rss-1c166df458de------2</link>
            <guid isPermaLink="false">https://medium.com/p/81c1040d488d</guid>
            <dc:creator><![CDATA[Pamornpol Jinachitra]]></dc:creator>
            <pubDate>Mon, 26 Jul 2021 03:19:20 GMT</pubDate>
            <atom:updated>2021-07-26T03:19:20.828Z</atom:updated>
            <content:encoded><![CDATA[<p>John Doerr เป็นนักลงทุนทางด้านเทคโนโลยีในตำนานคนหนึ่ง เขาเป็น partner ใน Sequoia Capital Venture Capital ที่ลงทุนในบริษัทชั้นนำตั้งแต่ยังเล็ก ไม่ว่าจะเป็น Google, Twitter, Amazon, Intuit etc.</p><p>OKR เหมือนเป็นเหล้าเก่าในขวดใหม่ที่ผสมผสาน Peter Drucker’s MBO (Management By Objectives), mainstream KPI มาจนถึงการนำไปใช้ใน INTEL ในยุค 90 มาถึงมือ John Doerr ก็นำไปแนะนำใช้ในบริษัท startup ทั้งหลายที่เขาเข้าลงทุน และที่เห็นผลชัดๆในกรณีของ Google ที่นำไปใช้อย่างสำเร็จ</p><p>ใจความสำคัญของ OKR คือ Objective และ Key Results</p><p>Objective คือ สิ่งที่จับต้องได้ เป็นรูปธรรม, เน้นการกระทำสิ่งที่จะทำให้ไปถึงจุดหมาย(ไกลๆ)ได้;</p><p>Key Results เป็นตัว benchmark วัดผลและเอาไว้ตรวจทานว่าเราจะไปถึง Objective ได้อย่างไร มันจะต้องเป็นการวัดที่จำเพาะ (specific) กำหนดด้วยเวลา (time-bound) และวัดได้ (measurable) และเป็นขั้นตอนที่เป็นเหตุเป็นผลกับการไปถึง Objective</p><p>ความได้ผลของ OKR มาจากหลัก (4 Superpowers)</p><ol><li>Prioritise and focus</li><li>Align and connect for teamwork</li><li>Track progress. Periodic check-ins. Continuous reassessment.</li><li>Stretch for amazing results</li></ol><p>ความสำคัญอื่นๆ ได้แก่ การที่มี alignment ทั้งองค์กร การพร้อมใจทำ action ไปในทิศทางเดียวกัน จะทำให้เกิดแรงที่มากกว่าการบวกกันธรรมดาๆ (more than the sum of its parts) เราอาจเคยได้ยินนิทานที่ว่าฝูงนกติดกับดักแห ต่างตัวต่างพยายามบินกันไปคนละทาง แต่ก็ยังติดอยู่ จนกระทั่งมีนกตัวหนึ่งที่เป็นผู้นำฝูง บอกให้ทุกตัวบินขึ้นในทางเดียวกัน แรงของทุกตัวที่ไปทางเดียวกันก็ทำให้ยกแหออกได้ (บางเรื่องอาจจะเป็นปลาแทน but you get the point)</p><p>การแยกการกำหนดเป้าหมายที่ดูห่างไกล เป็นการช่วยให้คนกล้าที่จะตั้ง โดยค่อย break down วิธีการไปให้ถึงตรงนั้นใน Key Results</p><p>การเขียนและประกาศให้ผู้อื่น ได้รับการพิสูจน์แล้วพิสูจน์อีกว่ามีผลทางจิตวิทยาทำให้คนที่เขียนหรือประกาศมีโอกาสที่จะ follow through ในสิ่งที่ประกาศไว้ (Quote: “92% said they would be more motivated to reach their goals if colleagues could see their progress”)</p><p>การเห็น OKR ของบริษัทจากผู้บริหารทำให้ Goal สามารถมาจาก Top Down ได้ด้วยเพราะพนักงานมีความเข้าใจ และสามารถเสนอการทำสิ่งใหม่ๆที่จะทำให้ company OKR เป็นจริง</p><p>OKR ที่ดีมักจะตั้งเพื่อให้ได้ 10x improvement และคาดหวังราว 60–70% ที่จะไปถึงตรงนั้น</p><p><strong>ต่างจาก KPI ยังไง?</strong></p><p>ถ้าว่ากันแบบง่ายๆ Key Results ก็มองว่าเป็น KPI อย่างที่ใช้กันอยู่ได้ ความต่างอื่นมาจากการที่มี Objective มาครอบ และการมองเห็น OKR ของแต่ละคนทั้งองค์กร ทำให้เกิด alignment ที่ดี อีกทั้ง OKR มักจะตั้งให้ไกล แบบไม่ต้องคิดกลัวไปไม่ถึง เพื่อให้ไม่ยึดติดกับ KPI ตัวเล็กตัวน้อย ที่มักเป็นการวัดผลในระยะสั้น OKR เหมาะกับองค์กรที่ต้องการโต 10x หรือมีการเปลี่ยนแปลงที่ใหญ่ได้อยู่เสมอ หรือเพิ่มความ agile (เพราะ KPI ที่ใช้กัน มักจะกำหนดเป็นรอบปี หรือสองครั้งต่อปี เพราะผูกกับค่าตอบแทนและโบนัส)</p><p>ทีนี้จะกล่าวถึงรายละเอียดในแต่ละ superpower ของ OKR</p><p><strong>Prioritise and Focus</strong></p><p>OKR ที่ดีต้องมีโฟกัส เริ่มจากการถามตัวเราเองว่า อะไรที่สำคัญในหกเดือน หนึ่งปี ข้างหน้า โดยเลือกมาเพียงไม่กี่อย่าง ที่สอดคล้องกับ Objective ของบริษัท OKR ที่ดีควรจะบ่งบอกหรืออธิบายได้ว่าต้องการทำอะไร (what) และทำไม (why)</p><p><strong>Alignment</strong></p><p>เมื่อพนักงานทุกคนทำทุกอย่างที่ไปในทิศทางเดียวกับวิสัยทัศน์ของบริษัท ผลรวมมักจะดีกว่าเป็นเท่าตัว แต่โดยทั่วไปทำได้ยาก พนักงานมักไม่ทราบหรือเข้าใจยุทธศาสตร์ของบริษัท และสิ่งที่คาดหวังจากพวกเขา OKR ที่ดีจึงต้องมีความโปร่งใส (transparent) เห็นได้จากทั้งบริษัท บางครั้ง Goal-setting จากพนักงานเอง (ตัวอย่างเช่น 20% free time ที่ใช้ที่ Google) ช่วยให้ alignment ไม่มากเกินไปจนทำให้พนักงานรู้สึกติดอยู่ในกรอบที่กำหนดด้วย OKR ที่ align กับ OKR ที่กำหนดโดยบริษัทอย่างเดียว</p><p><strong>Track</strong></p><p>เราสามารถติดตามผลของ OKR ได้ด้วยการเช็คอยู่เป็นระยะ (periodic check-ins) ให้เกรด (objective grading) และทบทวน (continuous re-assessment) ที่สำคัญ OKR ถูกปรับได้ตามสถานการณ์ที่อาจเปลี่ยนไป</p><p>OKR lifecycle เริ่มจากการเซ็ตอัพ (ซึ่งต้องอยู่ในลักษณะที่ทุกคนเห็นได้) เขียนเป็นลายลักษณ์อักษร ต่อจากนั้นต้องมีการ check-in เป็นระยะ การเขียนเป้าหมายและสิ่งที่ต้องกระทำลงเป็นตัวหนังสือเพิ่มความน่าจะเป็นที่มันจะเป็นจริงได้ การเช็คเป็นระยะๆยิ่งช่วยให้ความน่าจะเป็นนั้นเพิ่มขึ้นอีก ในแต่ละ check-in นั้น เราสามารถจะทำต่อไป หรือเปลี่ยน และทุกครั้ง(ไม่ว่าจะเปลี่ยนหรือไม่)สะท้อนกับตัวเองและทีมว่าเราได้เรียนรู้อะไรจากที่ผ่านมา เฟสสุดท้ายคือการให้คะแนน (objective scoring) เป็นเรื่องของการประเมินตัวเอง (self-assesment) การได้คะแนนต่ำเกินไป อาจหมายถึงต้องทบทวนว่า goal นั้นยังเหมาะสมหรือไม่ หรือต้องเปลี่ยนแปลงอะไรที่จะทำให้ดีขึ้นได้ ส่วนคะแนนเต็ม 100% ก็บ่งบอกว่า goal ที่ตั้งไว้อาจจะต่ำไป</p><p><strong>Stretch</strong></p><p>OKR สามารถกระตุ้นให้คนทำได้มากกว่าที่คิดว่าจะเป็นไปได้ การตั้ง goal ให้ไกลๆยากๆ (stretch goals) กระตุ้นให้คนออกจาก comfort zone ตัวอย่างที่ Google คือ ที่ Google จะมีการแบ่ง OKR เป็นสองกลุ่ม กลุ่มแรกคือ committed goals ที่ต้อง achieve ให้ได้ 100% เนื่องจากเกี่ยวข้องกับ product release หรือการจ้างงาน อีกกลุ่มเป็นเรื่องของ aspirational goals ที่มองภาพใหญ่กว่าและมีความเสี่ยงสูงที่จะไม่สำเร็จ (40% failure rate is expected) การมี goal ที่ไม่ต้องคาดหวังว่าจะต้องได้ 100% ทำให้เรากล้าเสี่ยงมากขึ้นเพราะจะสำเร็จแค่ 50%-60% ก็ไม่ได้เป็นความล้มเหลว ไม่มีอะไรต้องเป๊ะ ในขณะเดียวกันความเป็น stretch goal ทำให้เราไปมากกว่าที่ปกติทำ แล้วยังทำให้เรากระตือรือร้นอีกด้วย</p><p><strong>Continuous Performance Management</strong></p><p>อันนี้แถมจากเนื้อหาหลักของ OKR ในหนังสือ John Doerr บอกว่าแนวโน้มใหม่ของการบริหารจัดการผลงานพนักงานแบบใหม่คือการที่ทำการพูดคุยและประเมินอยู่ตลอดเวลา มากกว่าที่จะมีการกำหนดการประเมินผลงานประจำปีที่ใช้เวลามาก และขาด engagement ในระหว่างเวลานั้นๆ การมี feedback จากคนอื่นๆ ทีมอื่นๆ การพบปะพูดคุยเพื่อ check-in กับหัวหน้างาน รวมถึงการได้รับคำชมหรือ recognition จากผลการทำงานอยู่อย่างสม่ำเสมอทำให้พนักงานรู้สึกมีส่วนร่วมและต้องการที่จะอยู่กับบริษัทนานๆ</p><p>OKR เป็นการสร้างวัฒนธรรมองค์กรให้มีความรับผิดชอบต่อตนเอง (accountable) โปร่งใส (transparent) และเน้นการใช้วิสัยทัศน์ (vision-based) เป็นวัฒนธรรมที่เน้นการบอกผู้คนในองค์กรว่า<em>ควร</em>ทำอะไร แต่ไม่เหมือนการสร้างกฎที่บอกว่า<em>ต้อง</em>ทำอะไร ทำให้ทุกคนมีบริบทอยู่เสมอ ไม่ต้องให้ใครมาคอยกำกับ</p><p>ไม่น่าจะจบได้ดี ถ้าไม่มีตัวอย่างสักหน่อย (สมมุติเป้าหมายหลักคือ ตอนสิ้นปี บางครั้งละไว้ในฐานที่เข้าใจ)</p><p><strong>ตัวอย่างสำหรับ CEO (Note: there can be more than one Objectives)</strong></p><p>Objective 1: Grow our business to show sustained profitability</p><p>Key Results:</p><ol><li>Generate or have bookable 100M revenue for this fiscal year</li><li>Make recurring revenue 50M per year (from the above)</li><li>Launch 2 (scalable) products by end-of-year</li><li>Be able to close deals with 20 new customers by end-of-year</li></ol><p>Objective 2: Strengthen the team</p><p>Key Results:</p><ol><li>Find Tech Lead or Team Lead for dev</li><li>Increase head counts in dev team to 10</li><li>Roll out OKR program, knowledge sharing and biweekly activities</li><li>Find 3 more AI/ML engineers in critical area</li><li>Weekly training with Product Managers</li></ol><p><strong>ตัวอย่างสำหรับ Sales</strong></p><p>Objective: Grow revenue with new customers</p><p>Key Results</p><ol><li>สร้าง Qualified Lead ใหม่ให้ได้ 40 ราย ก่อนสิ้นปี (Comment: Note how it aligns with CEO’s objective above)</li><li>เพิ่ม Channel Partner จาก 2 เป็น 10</li><li>เพิ่ม ARR จาก new customers เป็น 40M (Comment: Note how it aligns with CEO’s objective above, maybe assuming, there is already ~8–10M ARR from existing customer.)</li></ol><p><strong>ตัวอย่างสำหรับ Product Manager</strong></p><p>Objective: Launch product by end-of-year</p><p>Key Results</p><ol><li>หา dedicated developer ประจำทีม 2 คน</li><li>Finalize features ที่จำเป็น (MVP) ต่อการ launch ภายในสิ้นเดือนสิงหาคม</li><li>Get 5 independent user feedback (both internal or external)</li></ol><p><strong>ตัวอย่างสำหรับ Application Developer</strong></p><p>Objective: Improve user experience using our AI-as-a-serivce</p><p>Key Results</p><ol><li>ทำเวลา onboarding process จาก 5 นาที ให้เหลือ 2 นาที</li><li>ทำ System Usability ให้ได้ score &gt;= 80</li><li>Get feedback score of 4 out of 5 on average between now and end-of-year</li></ol><p><strong>ตัวอย่างสำหรับ Machine Learning Engineer</strong></p><p>Objective: Improve user experience using my implementation of AI-as-a-service</p><p>Key Results</p><ol><li>เพิ่ม accuracy ของโมเด็ลจาก 90% เป็น 99% ด้วยการวัดจาก test bench X</li><li>เพิ่ม speed of response จาก 10 วินาที เป็น 3 วินาที</li><li>ทำ API ให้มี complaint จาก error หรือ user ต้องถามมาเกี่ยวกับวิธีใช้ ในส่วนที่เกี่ยวกับโมเด็ลเท่านั้น ไม่เกิน 1 ครั้ง หลังจาก production release จนถึงสิ้นปี</li></ol><p><strong>ตัวอย่างสำหรับ Marketing</strong></p><p>Objective: Make the company a reputable leader in AI</p><p>Key Results</p><ol><li>Launch เว็บไซต์ใหม่ที่มีผู้เข้าชม 1,000 views per month ด้วย bounce rate &lt; 50%</li><li>สร้าง content ที่ดึงดูด &gt; 10 views ทุกๆสัปดาห์</li><li>Publish ใน 1 advertorial ก่อนสิ้นปี</li><li>ทำให้ลิงค์ของบริษัทขึ้นมาใน Google Search 5 อันดับแรก เมื่อค้นหา Keyword ที่เกี่ยวข้องโดยตรง หรือที่เป็นชื่อบริษัท ชื่อ product</li><li>จัด 1 public event</li></ol><p><strong>ตัวอย่างสำหรับ HR</strong></p><p>Objective: สร้างทีม สร้างวัฒนธรรมองค์กร สร้างความสุข</p><p>Key Results</p><ol><li>Turn over rate &lt; 10%</li><li>Employee job satisfaction 4 out of 5</li><li>หา Recruiter เพิ่ม 2 เจ้าที่ specialize ด้าน IT</li><li>หา Job board หรือช่องทางที่จะได้ software developer ที่ดีได้ 2 ช่องทาง</li><li>หา Dev เพิ่มให้ได้ครบ 10 คน (Comment: Note how number 3,4,5 align with CEO goals above)</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=81c1040d488d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[According To Me, The Series: “The Man Who Solved The Market: How Jim Simons Launched the Quant…]]></title>
            <link>https://pamornpol.medium.com/according-to-me-the-series-the-man-who-solved-the-market-how-jim-simons-launched-the-quant-1fa7c3ad59c5?source=rss-1c166df458de------2</link>
            <guid isPermaLink="false">https://medium.com/p/1fa7c3ad59c5</guid>
            <dc:creator><![CDATA[Pamornpol Jinachitra]]></dc:creator>
            <pubDate>Sun, 28 Feb 2021 08:03:39 GMT</pubDate>
            <atom:updated>2021-02-28T08:03:39.849Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>According To Me, The Series: “The Man Who Solved The Market: How Jim Simons Launched the Quant Revolution” by Gregory Zuckerman</strong></h3><p>This is basically a story of Jim Simons and the rise of Renaissance Technology (RenTech), a hedge fund with nearly mystical reputation of achieving in excess of 60% return yearly since 1980. Obviously, such number implies he and all those involved must have become extremely rich. What is interesting is the story of his background, people he recruited and the dynamics of things at the firm and how it happened. For someone outside the financial world but of technical and academic background, his story is fascinating and suddenly becomes relatable. Of course, it is hard to imagine how each of us could do the same given less of a genius we are than he was. And for those hoping to find secret formula and spelled out steps to do so, you will be somewhat disappointed. RenTech’s algorithms remain as secretive as ever. But at least, given the background of each of his team, we can speculate what gave them the edge that made them so successful.</p><p>One of the ingredients for success seems to be Jim Simons’ own mathematical ingenuity. He already possessed a gifted mathematical ability from childhood, went through prestigious school studying maths, and was in academia for a long time before deciding to apply his maths ability to stock market. He started the fund Monemetrics at the age of 40! No small feat and what a conviction that he was able to convinced many similarly smart people to join in the endeavour (another lesson: surround yourself with someone smarter than you). One of them, Leonard Baum, was also in academia before joining RenTech (my discovery from this book that the inventor of <a href="https://en.wikipedia.org/wiki/Baum%E2%80%93Welch_algorithm">Baum-Welch algorithm</a> I used many times over during my Ph.D. study in the area of speech processing and others all over the world was an eye opening of how one can cross over to become so successful and profitable in another field using technical expertise applied in another).</p><p>The next ingredient is technology. The algorithms were the product of mathematical derivation and the minds behind it. However, it would not be practical without the use of computing technology to arrive at and deploy for execution in real time. RenTech seemed to have used technology well to gain an edge over others with automation of calculation and decision from data.</p><p>Another ingredient of success is conviction. Firstly, it was certainly not an easy decision to start a hedge fund at the age of 40 as mentioned. The belief in himself and other members of the team belief in what they are doing is certainly important. Secondly, there were many adversity along the way: the start from nothing, the people, the market crash, requires conviction to carry through. Also, even though apparently, a human intervention was employed controversially within the company when the stock market crash happened, mostly the team believed in the algorithms and let the maths do the trade with little intervention though the years.</p><p>Last ingredient worth mentioning is hard work. Nothing can replace that.</p><p>This book also talked about many colourful characters taking part along the way of RenTech. Some are simple tech nerds who enjoyed technical work. Some have big character and got involved in politics using his wealth as a means to influence political agenda. Some, Jim Simons himself, use their wealth in philanthropy. Just another angle of life outside work for these people for us to reflect on how we want our life and achievement to entail.</p><p>P.S. You don’t have to be right 100% of the time: “We’re right 50.75% of the time … but we’re 100% right 50.75% of the time” Robert Mercer</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1fa7c3ad59c5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[According to Me, The Series: “Subscribed” by Tien Tzuo]]></title>
            <link>https://pamornpol.medium.com/according-to-me-the-series-subscribed-by-tien-tzuo-b39f39c0678d?source=rss-1c166df458de------2</link>
            <guid isPermaLink="false">https://medium.com/p/b39f39c0678d</guid>
            <dc:creator><![CDATA[Pamornpol Jinachitra]]></dc:creator>
            <pubDate>Sun, 28 Feb 2021 08:03:11 GMT</pubDate>
            <atom:updated>2021-02-28T08:03:11.099Z</atom:updated>
            <content:encoded><![CDATA[<p>This book makes a convincing arguments of why businesses had moved or should move to subscription-based model and participate in the subscription economy albeit in the interest of the author’s founded company Zuora.</p><p>The way companies just design product to be released every year or so and hope it would be a hit is fading away. These days, ubiquitous Internet connection allows as-a-service to become possible. Coupled with the demand of “now”, with people not wanting to have an ownership of physical things, services where customers and their experiences take center stage, rather than product, become a new business model. Subscription model allows service to evolve over time all the time, with customers data pouring in and enabling technology like AI and data analytic. Instead of 1-year cycle of product iteration, the service can keep improving according to customers and market feedback. The real distinction between successful and failed business in the next decades is not online versus bricks-and-mortar, but more about who is being customer-centric (data-centric, customisation, flexible and multi-channel) which is now possible with subscription model versus old traditional ways of selling products.</p><p>Case in point being NY Times. With the onset of printed matters decline, NY Times was predicted to be doomed. However, after a successful execution of subscription service, its stock recovered from the nadir and skyrocketed ten-fold. Compared to ad-based business model that some digital news media follow like Buzzfeed or Mashable, NY Times created contents that customers find enough value to pay for (10% of customers are paying customers) and use the platform to expand internationally.</p><p>One attractive aspect of the subscription model is that it can give predictive revenue. Once you learn about and minimise your churn rate (rate at which your paying customers leave your service), you can predict with great certainty of what your revenue will look like next month from the base of your paying customers. New paying customers acquired will then add to this base revenue. This helps with planning and scaling operations much more easily instead of having a big hit or product or project but can never be sure of what will come next until the next product rolls out or next project’s contract signed. We are now talking about Customer Lifetime Value rather than one-off sale. Typically, subscription model loses money the first two years, but beyond that it is pure profit. But in order to reach that point, customer needs to be satisfied enough that they continue to subscribe and pay for the service. Investors also tend to value this predictability too, giving higher multiples to stock of SaaS companies than traditional ones.</p><p>Interestingly, many businesses that most may think cannot fit subscription economy by definitions actually have done it. The simple example being Apple’s iPhone. The smartphone market has come to the point of saturation. Consumers replace their phone with a new one at a longer cycle period. Innovation is also hard to come by to make one smartphone significantly better than the previous version. This is why Apple is now growing their services at a faster rate than their iPhone’s growth. These services are just leveraging on the ubiquity of iPhone products which now act as a conduit for consumers to experience Apple’s services like iCloud, iPhoto, Apple Music (which turned from 99 cent a song to unlimited number of song streaming with a flat subscription) etc.</p><p>More intriguing examples are guitar maker, Fender, whose sale of guitars declined. However, they started a subscription services to guitar lessons which helped drive people to want to buy their guitars more and stop people who bought their guitar from quitting. This is a good example of how a company can keep relationship, learn from their paying customers, even after the sale of a product. Caterpillar has already provided service of hauling dirt after using drones to make estimate of the area then charge by volume.</p><p>How to move from traditional model to subscription model is another topic discussed at length in this book. The case study is Adobe who moved their Creative Suite from one-off buying to a subscription model over 3 years. Inevitably, there were a number of quarters where revenue shrank significantly due to how subscription model pushes forward revenue that is yet to be realised in the future. They overly communicated to all stakeholders and did not force customers to switch immediately. They also had to rewrite financial models to factor in annual recurring revenue. This has become a blueprint for businesses who want to move from traditional to subscription model.</p><p>On the IT buyer side, the book suggested that OpEx is now preferred over CapEx. While CapEx allows companies to take advantage of amortisation and depreciation of the lump sum investment, OpEx requires less initial upfront investment and is more flexible. Not sure if that is widely accepted in Thailand but I believe this will also eventually happen. In my opinion, with OpEx that is lower than CapEx, companies keep the cash on hands which is definitely better than saying it is still on the book as “investment” or “asset” to be amortised over time whereas actual cash has already left the company’s bank account.</p><p>In order to be successful at Subscription model, the book suggests quite a few things. One that is stuck with me is about staying in beta forever. With the concept of Beta meaning it is still not a finished product, we should let it constantly evolve. The other is on IT. With subscription, the architecture of IT system should put a customer ID at the center stage with pricing, data and everything related connected to the ID rather than siloed into different departments (obviously, Zuora comes in nicely here to solve the problem).</p><p>In all, the book made quite a convincing argument of why and how to follow subscription model.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b39f39c0678d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[According To Me, The Series: “Never Split the Difference” by Chris Voss]]></title>
            <link>https://pamornpol.medium.com/according-to-me-the-series-never-split-the-difference-by-chris-voss-cb04a14290f1?source=rss-1c166df458de------2</link>
            <guid isPermaLink="false">https://medium.com/p/cb04a14290f1</guid>
            <dc:creator><![CDATA[Pamornpol Jinachitra]]></dc:creator>
            <pubDate>Sun, 28 Feb 2021 08:02:46 GMT</pubDate>
            <atom:updated>2021-02-28T08:02:46.384Z</atom:updated>
            <content:encoded><![CDATA[<p>เอาเป็นภาษาไทยสักครั้ง</p><p>หนังสือเล่มนี้น่าสนใจสำหรับคนที่อยากรู้ว่าคนเขา“ต่อรอง”กันอย่างไร เขียนโดยอดีตนักเจรจาต่อรองแห่งFBI ที่เชี่ยวชาญด้านการเจรจาต่อรองกับผู้ก่อการร้าย โจรลักพา ที่มีตัวประกัน (hostage negotiator) แต่ทุกสิ่งทุกอย่างมีพื้นฐานจาก human psychology และ behavioral science ซึ่งapply กับการต่อรองทางธุรกิจ กับคนในครอบครัว การค้าขาย และอื่นๆอีกมาก ถ้าจะสรุป Tips &amp; Tricks มาทั้งหมดต้องยาวแน่ๆ แต่สิ่งที่จำได้และประทับใจ (หลังจากกลับไปแอบอ่านสรุปของคนอื่นมาอีกที เพื่อเตือนความจำ บ่งบอกว่าจำอะไรไปใช้ไม่ค่อยได้เลย 😔) บวกมุมมองตัวเองนิดหน่อยก็ตามนี้</p><ol><li>ฟัง: ตั้งใจฟังโดยให้คู่สนทนาเป็นจุดแห่งความสนใจ เข้าอกเข้าใจความรู้สึกนึกคิดของอีกฝ่าย ทำให้คู่สนทนารู้สึกสบายใจ และปลอดภัยที่จะเปิดเผยความรู้สึกนึกคิดเกี่ยวกับหัวข้อหลักของการสนทนาต่อไป อันนี้ก็คล้ายกับ Culture Code เรื่องความรู้สึกปลอดภัยที่ต้องมาก่อนเรื่องอื่นๆ สร้างความรู้สึกปลอดภัย security เป็นพื้นฐานได้ เรื่องอื่นๆจะตามมาเอง</li><li>ใช้การ mirror ถามมาก็ถามกลับ ขออะไรมาก็พูดทวนกลับไป โดยปล่อยให้คู่สนทนาค่อยๆปล่อยข้อมูลจากฝั่งเขามาเอง เพื่อให้เราเข้าใจมากขึ้นว่าคู่ต่อรองเรามาจากไหนกันแน่ อันนี้ก็ตรงกับผนังสือเกี่ยวกับเทคนิคการขายเรื่องการ mirror ตามคู่สนทนาของเราเพื่อให้ sync กัน มนุษย์เรากลัวคนที่ต่างจากเรา แต่ชอบและสบายใจกับคนที่คล้ายเรา mirroring เป็น technique ที่ทำให้คนรู้สึกชอบเราได้ง่ายขึ้น</li><li>ให้ชื่อกับความรู้สึก (labeling) “ดูเหมือนว่าคุณรู้สึกโกรธที่…”เป็นทั้ง empathy ที่ แสดงว่าเราเข้าอกเข้าใจเขา และการ label ความรู้สึกทำให้คนเราจัดการมันได้ง่ายขึ้น</li><li>อย่าพอใจกับ yes คำตอบ noให้ข้อมูลเรามากกว่าเพื่อให้เห็นชัดขึ้นว่าอีกฝ่ายต้องการอะไร แต่ yes นั้นมีสามแบบซึ่งคล้ายกันและต้องพยายามแยกให้ออก 1) yes ที่เป็นการหลอกลวงหรือเป็นแค่คำตอบที่ง่ายๆให้หลุดจากสถานการณ์ อันนี้ต้องระวัง 2) yes ที่เป็น confirmation มักจะเกิดกับคำถามง่ายๆตรงไปตรงมา 3) yes แบบ commitment ซึ่งเป็นแบบที่เราต้องการมากที่สุด Noทำให้การเจรจา move forward เพราะมีข้อมูลและความชัดเจนมากขึ้น Noทำให้คนรู้สึกปลอดภัย in control และเราควรจะให้การปฏิเสธ No นี้เกิดขึ้นตั้งแต่ช่วงแรกของการเจรจา</li><li>ใช้การ paraphrase หรือสรุปความเข้าใจ เพื่อให้อีกฝ่ายบอกว่า “ใช่แล้ว” “That’s right” is better than a “yes”.</li><li>Anchor their emotions in loss aversion คนเรามักจะกลัวการสูญเสียอะไรบางอย่าง มากกว่าต้องการที่จะได้อะไร ปล่อยให้อีกฝ่าย anchor เรื่องตัวเงินหรือจำนวนก่อน เราแค่ฟัง</li><li>Establish a range โดยไม่ต้องพูดตัวเลขเป๊ะของเรา แต่อาจจะ quote range จากข้างนอก หรือ third party (แน่นอนว่าอันนี้ต้องทำการบ้านมาด้วย)</li><li>หลังจาก anchor ที่จำนวนตัวเลขสูงแล้วสามารถเปลี่ยนบางส่วนเป็นอะไรที่ไม่ใช่เงินได้ โดยเฉพาะสิ่งที่สำคัญกับเรามากกว่าอีกฝ่ายนึง</li><li>ใช้ตัวเลขละเอียด เช่น 124,595 มากกว่าจะ round out ไปเป็น 125,000 เพราะมันบ่งบอกว่าเราได้คิดมาดีแล้ว (อันนี้จำได้ไม่ยากและใช้มาตลอด ยังไงๆเวลาประเมินราคาก็ใช้สูตรเป็นตัวเลขเป๊ะๆอยู่แล้ว)</li><li>หลังจาก anchor ด้วยตัวเลขสูงแล้ว เราสามารถ offer surprising gift ที่ไม่เกี่ยวกับเรื่องที่กำลังเจรจากันอยู่เลยได้</li><li>สร้างภาพมายาคติว่าอีกฝ่ายมี control (The Illusion of Control) โดยการไม่บอกตรงๆว่าต้องการอะไร แต่ให้อีกฝ่ายแนะนำตัวเลือกมาสำหรับสิ่งที่เราต้องการคร่าวๆ และถามคำถาม calibrated question ที่ไม่สร้างความรู้สึก conflict แต่ช่วยguide ให้คู่เจรจาของเราคิดหาคำตอบด้วยตัวของเขาเอง (ไปในแนวทางที่เราต้องการ) calibrated questions เป็นพวก how? กับ what? ที่ต้องให้อีกฝ่ายอธิบาย หลีกเลี่ยงคำถามที่จะได้ yes หรือข้อมูลเล็กๆน้อยๆ เพราะเรามักจะเป็นฝ่ายที่ถูกคาดหวังให้ให้ข้อมูลอะไรกลับมาบ้างหลังจากนั้น อย่าถาม why? เพราะมักจะทำให้อีกฝ่ายรู้สึกว่าถูกกล่าวหา</li><li>เมื่อได้ yes แล้วต้องให้แน่ใจว่ามี plan of action/execution ด้วย “จะลองดู” ไม่ดีพอเพราะมันมักจะแปลว่ามันจะ fail (อันนี้เจอบ่อย เพราะเราเองก็ทำ ถ้าบอกจะลองดู ทำดีที่สุด มันมีแนวโน้มจะ fail หรือไม่ทำ ตั้งแต่แรก แต่ถ้าบอก ฉันจะรีวิวหนังสือ 5 เล่มตามนี้ มันก็จะเกิดขึ้น)</li></ol><p>นอกจากนั้นก็มี trick การจับโกหก วิธีตั้งตัวเลขแบบ Ackerman model ที่ช่วยการเจรจาต่อรอง ข้อ 1–3 เป็นข้อที่รู้สึกว่าต้องทำด้วยความจริงใจเท่านั้น ถึงจะเวิร์ค ไม่มีการ manipulate ใดๆทั้งสิ้น ใน step เหล่านี้ เพราะถ้าเรามัวแต่คิดคำนวณเรื่องการฟัง การเข้าอกเข้าใจ มันมักจะไม่ฟัง และก็ไม่ได้เข้าอกเข้าใจจริงๆ! ส่วนข้อหลังๆนี่ต้องฝึก ไม่น่าจะมาได้โดย soft skill อย่าง EQ</p><p>แม้ว่าการอ่านหนังสือเล่มนี้ ไม่ได้ทำให้เราเป็นนักเจรจาต่อรองที่เก่งในทันที (Case in point, Me!) เพราะทุกอย่างต้องอาศัยการฝึกฝน แต่ก็ทำให้เราสังเกตสิ่งเหล่านี้ได้ดีขึ้นจากเหล่านักขายที่เราอาจต้องเจอ และแม้จะไม่ได้ใช้ การได้เรียนรู้เกี่ยวกับ human psychology ที่เกี่ยวข้องก็เป็นสิ่งที่น่าสนใจ ให้ความรู้สึกสว่างในสมองได้มากมาย (Enlightened) 😇</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cb04a14290f1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[According to Me, The Series: “The Culture Code” by Daniel Coyle]]></title>
            <link>https://pamornpol.medium.com/according-to-me-the-series-the-culture-code-by-daniel-coyle-a61cce0bce4b?source=rss-1c166df458de------2</link>
            <guid isPermaLink="false">https://medium.com/p/a61cce0bce4b</guid>
            <dc:creator><![CDATA[Pamornpol Jinachitra]]></dc:creator>
            <pubDate>Sun, 28 Feb 2021 08:02:24 GMT</pubDate>
            <atom:updated>2021-02-28T08:02:24.277Z</atom:updated>
            <content:encoded><![CDATA[<p>This book is easy to follow and always intrigue you at every turn in unveiling the secret of such an abstract concept like culture. It starts by describing how a group of kindergartens working together can beat a group of MBA students at a certain group task and it sounds totally plausible. It all has to do with the team dynamics. Whereas, kindergartens simply do, express themselves honestly and transparently until the best action wins out, adults tend to do the politics, sizing each other up, being “เกรงใจ” etc. that hinder the progress.</p><p>To me, this book along with books on Netflix (“<a href="https://thailand.kinokuniya.com/bw/9781984881885">No Rules Rules</a>”) and Ray Dalio (“<a href="https://thailand.kinokuniya.com/bw/9781501124020">Principles</a>”) consistently say that transparency and honest feedback are important attributes of culture to create a thriving organisation (when more than two experts or successful precedences say the same thing, you can start to be certain it must be true and it must be working). Much has been heard about office politics and how unproductive it could be yet still happen. This leads to one important point in the book, “Safety”.</p><p>You feel safe when you feel you can be yourself and express your honest opinions among peer groups or to your leaders. When you feel comfortable, having the right chemistry with people around you, you feel safe. Fear is the most dominant primitive feeling human ever have. Once it takes over (in “unsafe” environment), you flee or spend energy and time worrying about potential insecurity. When you feel safe, then you can be productive and produce your best work since you don’t have to spend .</p><p>While the book never mentions “safety” in the economic sense, only the social aspect of it, I would note here that in terms of pay, this again is consistent with the concept of universal income where the idea is, give people enough income to live by (hence, security), and they will become more productive to the society at large. It is consistent with how most startups pay, enough that will take care of the basic needs, and then put everything else into equities for long-term future alignment with the startup. Again, when I see consistent evidences across multiple places, I take notice.</p><p>A similar concept to “Safety” or perhaps the consequence of it is “Belonging”. When people feel they belong to the group, they feel secured. The book demonstrated how a sports team that perform professionally, yet feels like a family can achieve something greater than others would have predicted. I think this will be more true or more apparent when the group face adversity and a set-back (the team lost after being so close to lifting the trophy, they regrouped and finally won the next year). It is this sense of belonging that will enable them to overcome obstacles.</p><p>Other concept of “Sense of Purpose” and “Common Goals” might be more obvious for a good team to have but also not so easy to establish. Lack of focus is something that may threaten this. Future uncertainties and things that may sway you off course are others. This is why some might say courage to follow through is more rare than vision.</p><p>Beyond this, it is clear culture is what people do. All can be explained by human psychology and instinctive nature. When you think about it, it is believable. Whether it happens to your organisation, it is up to everybody there but leaders will need to set an example and build environment that nurture it. Good news to me is that leader’s quality in creating such a good culture does not seem too far-fetch: many are just about being a decent human being!</p><p>Book summary by others can be found here.</p><p><a href="https://www.getstoryshots.com/books/the-culture-code-summary/">https://www.getstoryshots.com/books/the-culture-code-summary/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a61cce0bce4b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[According to Me, The Series: “Zero to One” by Peter Thiel]]></title>
            <link>https://pamornpol.medium.com/according-to-me-the-series-zero-to-one-by-peter-thiel-226e7081e66b?source=rss-1c166df458de------2</link>
            <guid isPermaLink="false">https://medium.com/p/226e7081e66b</guid>
            <dc:creator><![CDATA[Pamornpol Jinachitra]]></dc:creator>
            <pubDate>Sun, 28 Feb 2021 08:01:21 GMT</pubDate>
            <atom:updated>2021-02-28T08:01:21.483Z</atom:updated>
            <content:encoded><![CDATA[<p>Even before becoming a founding CEO of a technology-oriented subsidiary of a big corporation, I had read Zero to One maybe twice. It is short (relative to many other business books), thought provoking and insightful. Having come from the mouth of a person of intrigue and successful startup entrepreneur like Peter Thiel makes it especially interesting.</p><p>Despite constraints being put on our company which makes us not a typical startup and hence lessons from the book may not apply directly, many lessons can still be adopted. This is also a book taken to heart by the founding member of the board which made us on the same page somewhat: we are here to create a progress in the manner of 0 to 1 (vertical breakthrough), rather than 1 to N (horizontal expansion). This writing is in the context of creating this new technology company.</p><p>Here are a few points that stick with me in the context of our company</p><ol><li>Create a monopoly</li></ol><p>What we need</p><ul><li>Proprietary technology which is 10x better than any existing solution (Our positioning of technical expertise and potential access to unique data sets should enable us to achieve this. When it comes to existing solution, it is not only competitor offering similar technologies but other ways of doing things, for example, manual data entry)</li><li>Network effect (Not directly applicable but always possible if you think creatively, I believe!)</li><li>Economy of scale (This is at the heart of what we need to do. One word comes to mind: Automation)</li><li>Branding (Ok. We will be most admired AI company in Thailand etc.)</li></ul><p>2. How to build a monopoly</p><ul><li>Dominate niche market (Surely, working with big mother company means we can dominate? The challenge is will others accept solution from competitor’s close relative.)</li><li>Once dominate, then expand to adjacent market. (For us, once insurance back office is dominated, we will be the AI back-office for other businesses too).</li><li>Build partnership. Don’t disrupt. Examples come to mind PayPal-Visa, Google-AOL in the early days.</li><li>Make last great development in a market and reap benefit of mature ecosystem (Online insurance. Online banking. e-Commerce. Some are mature more than others)</li></ul><p>3. The Secret</p><p>Basically it is about finding a secret no others have found, whether it is the technology to do certain things, a vision of the future that nobody thought would be possible, or the believe that the majority would disagree upon. Courage (to stick through) is even harder to come by than the genius that come up with the ideas/technology.</p><p>These things influence my thinking on how we build a company. Obviously, even with the blueprints like this, challenges remain. Courage to stick with it and convince people; employees, partners, customers or even regulators, that the future can be different (the way we propose it to be) is probably the hardest one.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=226e7081e66b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>