<?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 Lyndsay Yerbic on Medium]]></title>
        <description><![CDATA[Stories by Lyndsay Yerbic on Medium]]></description>
        <link>https://medium.com/@lyndsayyerbic?source=rss-a1cf1385642d------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*O-CAdHa4qXJQL4Pf</url>
            <title>Stories by Lyndsay Yerbic on Medium</title>
            <link>https://medium.com/@lyndsayyerbic?source=rss-a1cf1385642d------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 08:11:53 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@lyndsayyerbic/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[Conversation with an Expert: A Practical Guide to Building an Incremental Multi-Agent AI System for…]]></title>
            <link>https://medium.com/google-cloud/conversation-with-an-expert-a-practical-guide-to-building-an-incremental-multi-agent-ai-system-for-0a75e4a8b3ed?source=rss-a1cf1385642d------2</link>
            <guid isPermaLink="false">https://medium.com/p/0a75e4a8b3ed</guid>
            <category><![CDATA[google-cloud-platform]]></category>
            <category><![CDATA[enterprise-architecture]]></category>
            <category><![CDATA[agents]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[prototyping]]></category>
            <dc:creator><![CDATA[Lyndsay Yerbic]]></dc:creator>
            <pubDate>Thu, 21 Aug 2025 12:01:47 GMT</pubDate>
            <atom:updated>2025-08-22T02:06:34.434Z</atom:updated>
            <content:encoded><![CDATA[<h3>Conversation with an Expert: A Practical Guide to Building an Incremental Multi-Agent AI System for Business Process Automation</h3><p><em>Lyndsay Yerbic, Head of Generative AI Field Solution Architecture, Pier Paolo Ippolito, Generative AI Field Solution Architect</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z-O7k_3d1Fp7mILTZOdIYw.png" /><figcaption><em>Pier Paolo Ippolito, Generative AI Field Solution Architect</em></figcaption></figure><p>The AI topic <em>du jour</em> is agentic automation for business processes. It’s no surprise. We all share the dream of building an army of intelligent, flexible digital workers to handle the tedious stuff, freeing us humans to tackle the more creative, high-value problems.</p><p>Large Language Models (LLMs) make this dream feel incredibly close. But as anyone who’s actually tried to build a production-grade agent will tell you, it isn’t magic. Building a truly useful AI system is a journey of careful planning and deliberate, step-by-step work — a lot like building any other piece of serious software.</p><p>To get a real-world perspective, I asked <a href="https://www.linkedin.com/in/pierpaolo28/">Pier</a>, a member of our GenAI Solutions Architect team, to share his playbook. He has spent years in the trenches turning messy manual workflows into slick, automated, multi-agent systems. What he shared was a practical, phased approach that anyone can use to move from idea to implementation.</p><p>He has graciously agreed to share his framework below.</p><p><em>(For a deeper dive into the theory, we highly recommend reading </em><a href="https://docs.google.com/document/d/1rsaK53T3Lg5KoGwvf8ukOUvbELRtH-V0LnOIFDxBryE/edit?tab=t.0#heading=h.pxcur8v2qagu"><em>Agentic Design Patterns</em></a><em>.)</em></p><h3>From Messy Workflows to Multi-Agent Systems: A 4-Phase Playbook</h3><p>The promise of agentic AI is transformative, but realizing that promise requires discipline. In my work as a GenAI Solutions Architect, I’ve seen many projects get stuck in “pilot purgatory” because they leap to a technical solution before deeply understanding the problem. The most successful initiatives are not about magic; they are about methodical engineering.</p><p>To guide this process, I rely on a four-phase playbook that moves from foundational discovery to a robust, value-driven system. It treats agent development not as a science experiment, but as a core engineering discipline.</p><h3>Phase 1: The Detective Work — Finding the Right Problem</h3><p>Your first job isn’t writing code. It’s to observe. Before you even think about AI, you have to become an expert on the human process you want to improve.</p><p>The goal here is to understand a manual workflow so intimately that its cracks and opportunities become obvious.</p><p>Book a workshop with the Subject Matter Expert (SME) — the person who <em>lives</em> this workflow every day. Then, just shadow them. Record everything: their screen, their clicks, the documents they open, even the sighs of frustration. You’re trying to create a perfect, high-fidelity map of how work gets done right now.</p><p>Once you have that map, I filter potential projects through three key questions:</p><ol><li><strong>What’s the real impact?</strong> Sure, saving time and money is great, but can this unlock new revenue streams or create a capability that doesn’t exist today?</li><li><strong>Is this actually doable?</strong> Be brutally honest about what’s possible with today’s technology. Your first project needs to be a clear win, not a wild science experiment.</li><li><strong>Can we do this again?</strong> Look for a solution or pattern you can reuse in other parts of the business. The best projects create a template for future success.</li></ol><p>You’re looking for the sweet spot: a project with high, measurable value but only moderate technical difficulty.</p><p>And here’s my most crucial piece of advice for this phase:</p><p><strong>Don’t just try to slap an AI on top of what you’re already doing. The most powerful solutions are ‘agentic-first.’ They rethink the problem from the ground up and ask, ‘How would a team of intelligent agents solve this?’ The answer is often completely different from the human process — and far more effective.</strong></p><h3>Phase 2: The Monolith — Build One Thing Well</h3><p>Once you’ve picked your problem, the temptation is to start architecting a massive, beautiful system. Resist that urge. Please. Just start with a single, simple component that proves the core idea works.</p><p>I call this the monolithic prototype. It isn’t meant to be pretty. It’s a quick-and-dirty test, often just a single, long prompt that tells one agent to perform a sequence of steps and gives it all the necessary tools. The only goal is to solve 80% of the core problem in the simplest way possible. You need to prove the approach is viable before you invest significant time and money.</p><p>At this stage, you face a choice: use a boilerplate (like an<a href="https://github.com/GoogleCloudPlatform/agent-starter-pack"> agent starter kit</a>) or build from scratch. Honestly, if your task is fairly self-contained and doesn’t need deep cloud integration, starting from scratch is often cleaner. You avoid inheriting a bunch of code you don’t need.</p><p>This is also the time to start sketching out the future. I’m a big fan of using a mind map or a digital whiteboard (like Excalidraw). Put the main goal in the middle. Branch out with the absolute must-have capabilities for your prototype. Then, on totally separate branches, brainstorm all the ‘nice-to-have’ features — a news search, financial analysis, whatever. Write them down, and then park them. You’ll come back to them, I promise.</p><h3>Phase 3: The Assembly Line — From One Agent to a Team</h3><p>Eventually, your brilliant, do-it-all monolithic agent will start to stumble.</p><p>It’s like a star employee who’s been given way too many jobs. The prompt gets too long, and the agent starts forgetting instructions or getting confused. That’s your cue. It’s time to bring in the specialists.</p><p>This is where you break down your monolith into a “microservices-style” system of specialized agents that work together. The trick is to find the “natural division of labor.” Think about the distinct skills required. You might have a <strong>Data Scraper Agent</strong> that pulls info from APIs, a <strong>Fact Checker Agent</strong> that verifies it, and a <strong>Writer Agent</strong> that turns it all into a clean report. Each one is an expert at its single task.</p><p>So, how do you organize this new digital team? I rely on a few common patterns:</p><ul><li><strong>For step-by-step tasks,</strong> a simple <strong>assembly line</strong> is perfect. Agent A does its job, then hands its work to Agent B, and so on.</li><li><strong>For complex decisions,</strong> a <strong>hierarchy</strong> is better. An agent trying to choose between 50 different user intents will fail. Instead, have a main <strong>Orchestrator Agent</strong> make one big decision — like routing a request to either the ‘Billing’ team or the ‘Support’ team. That sub-agent can then handle the more specific choices. You want to build deep, not wide.</li><li><strong>For repetitive work,</strong> you can build <strong>loops</strong>. An agent can keep refining a draft or searching for data until it hits a certain quality bar.</li><li><strong>For gathering info,</strong> you can have agents work in <strong>parallel</strong>. If you need data from three different sources, three agents can fetch it simultaneously, then pass their findings to a fourth agent to synthesize the results.</li></ul><p>When you build this team, you have to ensure they can communicate effectively. Use the right tool for the job. Your orchestrator might be a powerful, top-tier model like Gemini 2.5 Pro, while the specialists can be smaller, faster models. Crucially, you must give them a shared memory — a common scratchpad where one agent can leave notes for the others.</p><p>One final, often overlooked, tip for orchestrators: you have to state the obvious. An LLM has no idea what day it is. If you need it to search for recent news, your prompt must literally begin with, “For context, today’s date is 20th of August 2025”</p><p>Once this core team is working smoothly, you can finally pull out that old mind map and start adding the “nice-to-have” features, knowing you have a solid foundation to build on.</p><h3>Phase 4: The Reality Check — Does It Actually Work?</h3><p>So, you’ve built it. Now for the million-dollar question: Is it any good?</p><p>How you evaluate your system depends on the project. For a simple internal tool, testing it against a checklist of examples might be enough. But for a customer-facing system, you need to be much more rigorous.</p><p>For those bigger systems, I suggest more advanced methods. You can generate synthetic data to test edge cases, or even use an ‘actor-critic’ pattern, where a second AI agent’s only job is to fact-check and grade the first one. To make this easier, you can use the evaluation utilities in frameworks like the Google Agent Development Kit to run structured tests and track performance over time.</p><p>But the ultimate test is, and always will be, human.</p><p>Before you pop the champagne, give the tool to the SME you shadowed way back in Phase 1. Let them use it, break it, and give you honest feedback. That’s the one insight you can’t get from any automated test. Does this actually make their day easier? Does it solve the <em>real</em> problem?</p><p>If the answer is yes, then you’ve done it. You haven’t just built a piece of tech; you’ve built a real solution. And that’s where the journey truly begins.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0a75e4a8b3ed" width="1" height="1" alt=""><hr><p><a href="https://medium.com/google-cloud/conversation-with-an-expert-a-practical-guide-to-building-an-incremental-multi-agent-ai-system-for-0a75e4a8b3ed">Conversation with an Expert: A Practical Guide to Building an Incremental Multi-Agent AI System for…</a> was originally published in <a href="https://medium.com/google-cloud">Google Cloud - Community</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Generative AI Design Patterns: Lessons from the Enterprise]]></title>
            <link>https://medium.com/google-cloud/generative-ai-design-patterns-lessons-from-the-enterprise-f86b10fde8f3?source=rss-a1cf1385642d------2</link>
            <guid isPermaLink="false">https://medium.com/p/f86b10fde8f3</guid>
            <category><![CDATA[enterprise-architecture]]></category>
            <category><![CDATA[google-cloud-platform]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[generative-ai]]></category>
            <category><![CDATA[best-practices-technology]]></category>
            <dc:creator><![CDATA[Lyndsay Yerbic]]></dc:creator>
            <pubDate>Wed, 02 Jul 2025 09:20:45 GMT</pubDate>
            <atom:updated>2025-07-03T01:38:59.375Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/942/1*aGwucdNKT1G7Vkc7dWE7wQ.png" /></figure><p>I recently had the opportunity to talk about generative AI design patterns to a group of non-profit organizations at a <a href="http://google.org/">Google.org</a> event, sharing insights from my work at the enterprise architecture level. It was an incredible conversation that triggered an ‘aha’ moment for me. In the tech industry, we may think the fundamentals (prompting, tuning, RAG) are table stakes knowledge of generative AI at this point, but the guidepost is still moving rapidly in terms of application best practices.</p><p>To confidently get started, sometimes <a href="https://medium.com/google-cloud/who-are-the-google-cloud-generative-ai-field-solution-architects-fsas-7541c425c35d">you need a guide</a> to talk through what’s “just right” for your use case and the tradeoffs. When faced with large decisions with far-reaching impacts, the right answer is not always clear. It’s hard to sift through the mass of information available to get specific and articulated advice (although Gemini is a great resource for this!).</p><p>I’ve distilled down the key takeaways of my full talk here.</p><h3>Why start with the “art of the possible”?</h3><p>It’s important to define your North Star (see my <a href="https://medium.com/google-cloud/got-100-ai-ideas-heres-the-framework-for-finding-the-gold-4c2a9efe891e">previous post</a> for a framework) when you’re choosing where to start.</p><p>Take these 2 examples:</p><ol><li>The new Application Build feature in <a href="https://aistudio.google.com/apps">AI Studio</a> (text -&gt; code -&gt; deployed app in minutes)</li><li><a href="https://www.youtube.com/watch?v=nXVvvRhiGjI">Project Astra</a> (now available to all through <a href="https://gemini.google/overview/gemini-live/?hl=en">Gemini Live</a>)</li></ol><p><strong><em>Less than a year ago, these were visions in a lab, and now they’re available to the world.</em></strong></p><p>Having this vision ensures that what you’re building and architecting today will scale to support the ultimate goal with the least rework and tech debt possible. Don’t drown in complexity — start small, break it into manageable chunks and make sure to set up for rapid experimentation.</p><h3>Foundational Generative AI Design Patterns</h3><p>Let’s dive into some foundational generative AI design patterns you can start using today. You can follow along — pick a topic and go to Gemini or AI Studio to test these prompts to demonstrate the patterns!</p><h3>Prompting — i.e. talking to the model in natural language</h3><p>Frontier LLMs today are trained on extensive data and are incredibly powerful. Coupled with long context windows, these LLMs negate the need for fine-tuning or RAG in some cases — just by crafting powerful prompts. Before investing in more complex patterns, <strong>always</strong> try the following prompting methodologies to see if you can get the results you want.</p><p><strong>When to use prompting only:</strong></p><p>Information needed is general. No engineering required to get the desired output from a foundational LLM.</p><ul><li><strong>Zero-shot prompting:</strong> Asking the model to perform a task without any examples.</li><li><strong>In-Context Learning: </strong>Providing examples of style, desired outcomes within the prompt, such as few-shot</li><li><strong>Chain-of-Thought (CoT) prompting:</strong> Breaking down complex problems into smaller, logical steps for the model to follow.</li><li><strong>Role prompting:</strong> Instructing the AI to adopt a specific persona or role.</li></ul><p>See more <a href="https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/prompt-design-strategies">advanced prompting techniques</a>.</p><p><strong>💡 Pro Tips:</strong></p><ul><li>Be specific with your prompts, avoid open ended questions</li><li>Add contextual information</li><li>Add more examples and sample patterns to the prompt to improve accuracy</li><li>Be careful with math and logic problems</li><li>Have multiple engineers try different prompts (I love this one)</li></ul><p><em>I was testing a demo app, and having trouble getting fast and clear results back. A colleague (the awesome Zack Akil) told me “You’re being too polite, be more direct. No fluff words.”</em></p><p><em>Think about this, when you prompt and LLM “can you do XYZ?”, the model goes through more processing: “Can I..?” → “Yes, I can..” → “Should I..?” → “Yes, assume that’s what user wants…” → </em><strong><em>Execute</em></strong><em>.</em></p><h3>Retrieval Augmented Generation (RAG) — i.e. give the LLM another brain</h3><p>For tasks that require access to external, up-to-date, or verifiable information, RAG is often a strong choice, either on its own or combined with prompting. When to use RAG:</p><ul><li>Information needed is not present in the model’s pre-trained knowledge</li><li>Fresher data — Grounding on Google Search is a good example of deploying RAG to keep results current</li><li>Mitigate hallucinations</li><li>Provide traceability and citations for where information is sourced</li><li>Gain efficiencies — reduce the need for models with very large context windows</li><li><strong>An LLM alone will not differentiate you </strong>— Everyone has access to the same LLMs. You differentiate by using your proprietary knowledge and data to come up with new insights.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/953/0*fkJ4sGhAEUKrjRs4" /></figure><p><strong>💡Pro Tip:</strong></p><p>Retrieval design and strategy is critically<strong> </strong>important to your RAG success. Chunking breaks large documents into manageable pieces to fit LLM context windows and improves retrieval efficiency. Ranking reorders the pieces to give the most relevant information first to the LLM. Be sure to <a href="https://cloud.google.com/generative-ai-app-builder/docs/parse-chunk-documents">read up on</a> chunking and advanced ranking techniques.</p><h3>Fine Tuning &amp; Advanced Mechanisms for Accuracy</h3><p>Creating a version of the model that is trained or tuned on specialised dataset for increased accuracy</p><p><strong>When to use these advanced methods?</strong></p><p>The LLM doesn’t have enough data or context to a specific process, terminology to get sufficiently accurate results</p><ul><li>High degree of specificity and domain expertise</li><li>Improved accuracy and consistency</li><li>Adapting to specific tasks or styles</li><li>Insufficient general knowledge</li><li>High-volume, repetitive tasks</li><li>Reduced inference latency</li></ul><p><strong>What methods are available to me?</strong></p><ul><li>Supervised Tuning — LLM with <a href="https://cloud.google.com/vertex-ai/generative-ai/docs/models/gemini-supervised-tuning">task-specific tuned parameters</a></li><li>Reinforcement Learning with Human Feedback — <a href="https://cloud.google.com/blog/products/ai-machine-learning/rlhf-on-google-cloud?e=48754805">reward driven assessment</a></li><li>Distillation step-by-step — here’s a <a href="https://research.google/blog/distilling-step-by-step-outperforming-larger-language-models-with-less-training-data-and-smaller-model-sizes/">DeepMind research blog</a> and <a href="https://colab.sandbox.google.com/github/GoogleCloudPlatform/vertex-ai-samples/blob/main/notebooks/official/generative_ai/distillation.ipynb">Colab</a> to get you started</li><li>Full fine tuning — <a href="https://cloud.google.com/vertex-ai/generative-ai/docs/models/tune-models">Introduction to Tuning</a></li></ul><p><strong>💡Pro Tip:</strong></p><p>When considering fine tuning or advanced techniques, think about smaller or open source models that may already be trained in specific areas. For example, the <a href="https://deepmind.google/models/gemma/">Gemma</a> suite is a good place to start to help you optimize time and costs.</p><h3>Decision Matrix</h3><p>My talk also covered tool use, agents, evaluation and production best practices. Way too much for one article, so stay tuned for part two! For now, I’ll leave you with a matrix to help you make well-informed generative AI decisions. Happy architecting!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IzY7RxkRGXUHc6ywwqpQ5A.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f86b10fde8f3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/google-cloud/generative-ai-design-patterns-lessons-from-the-enterprise-f86b10fde8f3">Generative AI Design Patterns: Lessons from the Enterprise</a> was originally published in <a href="https://medium.com/google-cloud">Google Cloud - Community</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Got 100 AI Ideas? Here’s the Framework for Finding the Gold]]></title>
            <link>https://medium.com/google-cloud/got-100-ai-ideas-heres-the-framework-for-finding-the-gold-4c2a9efe891e?source=rss-a1cf1385642d------2</link>
            <guid isPermaLink="false">https://medium.com/p/4c2a9efe891e</guid>
            <category><![CDATA[value-creation]]></category>
            <category><![CDATA[google-cloud-platform]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[generative-ai-consulting]]></category>
            <category><![CDATA[generative-ai-use-cases]]></category>
            <dc:creator><![CDATA[Lyndsay Yerbic]]></dc:creator>
            <pubDate>Sat, 29 Mar 2025 10:22:11 GMT</pubDate>
            <atom:updated>2025-03-31T05:22:52.104Z</atom:updated>
            <content:encoded><![CDATA[<p>In this blog series, I’m sharing practical tools, frameworks and lessons learned for accelerating successful AI projects. The first of this series is<strong> </strong>to help your team get from the initial ideation phase to a clear priority of projects and roadmap.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1SU8My4r-dmatCSkUmXnUA.png" /></figure><p>In our world, we hear a lot of innovative ideas on how to use generative AI. Check out my first blog for more info: <a href="https://medium.com/google-cloud/who-are-the-google-cloud-generative-ai-field-solution-architects-fsas-7541c425c35d">Who Are the Google Cloud Generative AI Field Solution Architects?</a> Generative AI has immense potential, but navigating the flood of possibilities can be overwhelming. Many customers I work with have identified dozens, even hundreds, of potential use cases, only to stall when deciding where — and how — to begin.</p><p>This post introduces a practical framework to get you started.</p><p><strong>1) Find Your “Why” with the Best, Better, Good Methodology</strong></p><p>While many frameworks exist, the key is identifying what’s most important and why — making sure to the right stakeholders in the process. Here’s a simple method to help.</p><p>I created the Best, Better, Good Methodology to nail down the core mission of what we are trying to accomplish, so we don’t get lost in fuzzy goals, lack of shared vision, and constraints too early. I find that it’s most effective thinking on a 2 year time horizon, with the ideas building a foundation for what could follow. Start with Best, because it focuses on the ultimate change you want to drive before identifying limitations.</p><p><strong><em>Best (The Dream):</em></strong><em> </em>Starting with the end user in mind, think about the best-case scenario. What does that look like, what elements does it include? This forces ambition and clarifies the North Star. Work back from here and identify critical blockers in accomplishing this. Note these potential blockers, and prioritize investigating their feasibility early on to understand the true constraints. <em>This gives you the vision.</em></p><p><strong><em>Better (The Strategic Tradeoff):</em></strong><em> </em>If your Best scenario has roadblocks that are insurmountable, move to Better. What in Best could you pragmatically sacrifice while still achieving significant impact? What about Best is not doable or requires too many tradeoffs, and are there high-value compromises? Follow the same due diligence as the Best scenario. <em>This clarifies your core priorities.</em></p><p><strong><em>Good (The Must-Have Core):</em></strong><em> </em>Repeat the same process for Best and Better. What’s the essential, must-have baseline? If forced to strip back further, what <em>must</em> remain to deliver core value? <em>This reveals your non-negotiables.</em></p><p>This structured cascade doesn’t just produce a list — it stress-tests your vision. It illuminates the consistent, non-negotiable elements across all three levels. <strong>That’s your true ‘Why’ — the core value proposition to build upon.</strong> With this clarity, you’re ready to frame the conversation around specific AI opportunities.</p><p><strong>2) Frame the Conversation Correctly</strong></p><p>From the context of your Best, Better, Good exercise, start your project ideation process. There are two critical elements in which to frame your brainstorming.</p><ul><li>What is most painful and important for us today? This could be an experience your customers hate, a competitive offering that is taking away marketshare, or tedious internal tasks that suck up valuable employee time.</li><li>Where is generative AI best fit for use cases?</li></ul><p>Let’s dig into the second point more. Generative AI can do a lot of amazing things, and gets better everyday, but there are areas where the effort is currently too high or the system complexity will jeopardize your pilot in the first phases with no results.</p><p>Frame your thinking about where generative AI fits in a few categories:</p><ul><li>Do we have data that will provide additional value beyond the foundational model’s general knowledge? If yes, you are strategically positioned to add value to a process.</li><li>Can generative AI automate answering frequently asked questions currently handled manually at scale? If so, even if the accuracy of a foundational model doesn’t answer with total precision today, this is a great area to focus because of the multitude of prompt engineering, training and tuning options available.</li><li>Creative work, generating new content and images/video</li><li>Searching across lots of information, summarizing insights, and finding specific points quickly</li></ul><p><strong>3) Prioritize: Business Impact to Complexity Mapping</strong></p><p>Out of this exercise, you should now have a sense of a) what is truly important and b) technical feasibility. These become your work use cases. Write them all out.</p><p>Next, map the use cases to a Business Impact to Complexity chart. This is a simple framework that helps us prioritize the roadmap. I’ve put a slight spin on it here, for generative AI use cases.</p><p><em>Map the use cases to a chart with Business Impact (Low to High) on the Y-axis and Implementation Complexity (Low to High) on the X-axis.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*R1RtuaqwPc8k61oY" /></figure><p>For example, for my dream Dog Wearables project:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*BDFlIooZUr60sjCU" /></figure><p>In this scenario, I’d start with the Remote Treat Dispenser (low complexity, high impact) to prove a quick win and learn valuable lessons. This initial project might involve developing sensor integrations or user interaction patterns via an app, creating foundational components that will eventually help build toward the Bark Translator (high complexity, high impact).</p><p><strong>Questions to ask during the mapping exercise:</strong></p><ul><li>Do we have the necessary, accessible, and high-quality data required for this specific use case?</li><li>Can this use case standalone, or does it require extensive reliance on other systems and integrations to be able to pilot successfully?</li><li>Is this a problem generative AI can/should solve, or should we consider other methods?</li><li>Is this measurable? How <em>specifically</em> are we defining business impact (e.g., KPIs like % time saved, error rate reduction, revenue generated)? Does this impact align to our “Best” goal?</li><li>Who will do this? Do we have the skillset in-house, do we need additional expertise?</li></ul><p>While technical deep dives, governance, and change management are crucial next steps, this framework provides the essential starting point. By systematically identifying your ‘Why,’ framing the potential accurately, and prioritizing strategically, you can cut through the noise and build a focused, actionable AI roadmap.</p><p><strong>Please leave comments, questions, tell me where you’re struggling with AI projects, or recommend future post topics!</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4c2a9efe891e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/google-cloud/got-100-ai-ideas-heres-the-framework-for-finding-the-gold-4c2a9efe891e">Got 100 AI Ideas? Here’s the Framework for Finding the Gold</a> was originally published in <a href="https://medium.com/google-cloud">Google Cloud - Community</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Who are the Google Cloud Generative AI Field Solution Architects (FSAs)?]]></title>
            <link>https://medium.com/google-cloud/who-are-the-google-cloud-generative-ai-field-solution-architects-fsas-7541c425c35d?source=rss-a1cf1385642d------2</link>
            <guid isPermaLink="false">https://medium.com/p/7541c425c35d</guid>
            <category><![CDATA[gemini]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[vertex-ai]]></category>
            <category><![CDATA[google-cloud-platform]]></category>
            <category><![CDATA[enterprise-technology]]></category>
            <dc:creator><![CDATA[Lyndsay Yerbic]]></dc:creator>
            <pubDate>Mon, 24 Mar 2025 17:25:26 GMT</pubDate>
            <atom:updated>2025-03-29T10:24:15.665Z</atom:updated>
            <content:encoded><![CDATA[<p>I get asked a lot: <em>“What does your team do?” </em>The answer is relevant to anyone working in the AI space today, from developers to executives.</p><p>The Google Cloud Generative AI Field Solution Architects (FSAs) have been in the trenches, turning AI ideas into real-world impact. I’m starting a series to share the stories, strategies, and the team that makes it happen. Follow me for frequent insights on <a href="https://www.linkedin.com/in/lyndsay-yerbic-616b671b/">LinkedIn</a> and <a href="https://medium.com/@lyndsayyerbic">Medium</a>.</p><p>Read on for real-world examples, and let me know your feedback in the comments on what you’d like to see in this series!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*7RtS7J3LO8U4v5M_L4zG1Q.png" /><figcaption>Image generated by Imagen, content written by me</figcaption></figure><p><strong>Who are the Google Cloud Generative AI Field Solution Architects (FSAs)?</strong></p><p>In 2024, AI leaders at Google Cloud built a new team. With the explosion of AI use cases and demand across industries, we saw the need to provide fast hands-on technical expertise to customers beginning the journey, or stuck on difficult concepts and problems.</p><p>We live in a world where we’re all trying to “see around the curve” of what’s next with AI. The FSA team has deep hands-on AI knowledge, and also understands that there are fundamental elements (such as data, operational foundations, evaluations) needed to build impactful solutions that can scale. We’re helping customers successfully pilot these projects, with best practices working with Google Cloud’s Gemini, Vertex, and beyond.</p><p>I could do a whole profile on the amazing people on the team — from data scientists to LLM experts — the FSAs are builders, coders, experimenters with passion for building safe and beneficial AI pilots for all. We take insights and best practices and share them, to accelerate time to AI reality and lower the barrier to entry.</p><p><strong>How we work</strong></p><p>Simply put, we build AI prototypes with customers. More on <strong>how</strong> we do this in future posts. As a teaser for upcoming posts, customers come to us with big ideas about how things will work in a new AI-driven world — from revamping traditional processes, to completely reinventing experiences. We dig into requirements and ideate on solutions and architectures. Many days we look at each other and ask “Can this be done?” And then we go build it, and discover ways to make it possible. We provide enablement and technical handovers to make sure teams are well-equipped to carry the work forward.</p><p>We work rapidly, and focus on solving the hardest AI problems facing organizations today.</p><p><strong>Real-world examples of global FSA work:</strong></p><p><a href="https://venturebeat.com/games/x-games-will-use-ai-tech-from-google-cloud/">X Games</a> — An AI-powered system for real-time athlete performance analysis.</p><p><a href="https://www.linkedin.com/posts/google-cloud_what-happens-when-gemini-analyzes-your-golf-activity-7272354642938707968-eN7Q?utm_source=share&amp;utm_medium=member_desktop&amp;rcm=ACoAAARDzLIBPCi8gUfYuCx_8_4PGOBOcB0JRU4">Golf Swing Analysis</a> — Personalized feedback on golf swings using computer vision and AI.</p><p><a href="https://www.linkedin.com/pulse/what-we-learned-from-co-building-genai-products-google-shezan-kazi-xpuze/?trackingId=cd%2Fw5eKfQoOCoJF4PnfKXw%3D%3D">What we learned from co-building genAI products with Google.</a></p><p><a href="https://medium.com/google-cloud/genai-app-starter-pack-now-with-rag-pattern-vertex-ai-search-fb81bf61bcd5">GenAI App Starter Pack — Now with RAG Pattern &amp; Vertex AI Search!</a></p><p><a href="https://medium.com/google-cloud/unlocking-customer-insights-with-ai-market-research-digital-twins-5c576ab3f184">Unlocking Customer Insights with AI Market Research Digital Twins</a></p><p><a href="https://www.prnewswire.com/news-releases/deutsche-telekom-and-google-cloud-partner-on-agentic-ai-for-autonomous-networks-302384349.html">Deutsche Telekom RAN Guardian Agentic AI</a></p><p><a href="https://cloud.google.com/blog/products/ai-machine-learning/how-signal-iduna-supercharges-customer-service-with-gen-ai?hl=en">Google Cloud Blog on Signal Iduna Health Insurance RAG Chatbot</a></p><p><strong>How to work with us</strong></p><p>First, follow us on LinkedIn and Medium! We’ll bring you real-time experiment findings, as well as our reflections and advice on best practices. Engage with us through the comments!</p><p>If you want to work directly with us on an idea, reach out to your Google Cloud representative for more information.</p><p>Stay tuned for next week, where I’ll be publishing Episode 1 of my series —</p><p><a href="https://medium.com/@lyndsayyerbic/got-100-ai-ideas-heres-the-framework-for-finding-the-gold-4c2a9efe891e">Got 100 AI Ideas? Here’s the Framework for Finding the Gold</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7541c425c35d" width="1" height="1" alt=""><hr><p><a href="https://medium.com/google-cloud/who-are-the-google-cloud-generative-ai-field-solution-architects-fsas-7541c425c35d">Who are the Google Cloud Generative AI Field Solution Architects (FSAs)?</a> was originally published in <a href="https://medium.com/google-cloud">Google Cloud - Community</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>