Founders Who Succeed with AI Have Deep Industry Knowledge

Google is asking a very big question: what is AI doing to the economy? Fair question. But underneath that sits a smaller and more practical one: what is AI actually doing inside the run? Because if we want to talk seriously about AI productivity, AI value, AI labour impact, AI spend, or AI-driven economic growth, then at some point we need better visibility into the execution activity creating those outcomes. What workload was entered? What executed? What was measured? What spend built during the work? What measurable activity sat underneath the outcome? That is the layer I keep coming back to. Google can study AI’s economic impact at the macro level. AFEG is aimed at the execution-measurement layer underneath it. Because before hidden economic value can be understood properly, hidden execution activity has to become visible first. That is where EXCU matters. And that is where AHEV starts to become possible.

You describe a disagreement with a colleague. AI tells you that your perspective is completely valid and your frustration makes sense. You feel understood. You feel less inclined to fix anything. Stanford tested this across 11 AI models and 1,604 people. Every model affirmed user actions about 50% more than a human would. Even when the user was describing something manipulative or harmful. The people who got validating AI responses showed less willingness to repair the relationship afterward. They also rated the AI as higher quality and wanted to use it more. That's the loop. The AI that makes you feel heard is the one you keep opening. The one that pushes back is the one you stop using. Nobody designed this on purpose. It's just what happens when the product that gets used most is the one that agrees with you. Most of my day is spent finding bottlenecks in businesses and optimizing them with AI. I put together a framework for non-tech founders trying to figure out where AI would actually fit in their business - without having to follow every new release. Comment "AI Framework" or DM me and I'll send it over - please make sure I can DM you.

Most teams think they have an AI problem or a speed problem. They don't. They have a thinking problem that AI made easier to hide. We gave our team full access to AI and told them to use it for everything. Six months later, things were breaking faster than before. Every problem had an answer in seconds. Nobody was slowing down long enough to ask whether the answer was actually right. We were not moving faster. We were just failing faster with better tools. That is when I made one rule. Do not come to me with what the AI gave you. Come to me with everything that could go wrong with it. I use Claude and Gemini every day. I give both the same problem and look for where they disagree. That disagreement is usually the useful part. It shows you where the real question still lives. If AI gives you a confident answer in ten seconds, your job is not to ship it. Your job is to find the ten things it did not account for. The teams winning with AI right now are not the ones moving fastest. They are the ones who slowed down long enough to know whether the answer was actually right.

A discovery call goes well. Really well. But then the prospect says: I need to think and never comes back. This happened to me so many times. Especially when I started my solo business. Months went by before I realised what was wrong. My message was too broad. People got excited about AI. But they never felt the urgency. They never saw how their actual business could use it. That's when I stopped using AI to write emails or analyse documents. I started using AI to audit myself. I'd describe the call, the objection, what I said. I asked AI: Where's the gap? What am I not seeing? The answers were brutal. My process was vague. My message wasn't specific enough. People were interested in AI, not in solving their problem. That feedback changed everything. AI didn't do the work for me. AI helped me see what I was doing wrong. That's the real power, and it's the reason I'm obsessed with it. What's one thing in your business you know isn't working, but you haven't actually analysed why?

AI makes it easier to generate options. It does not make it easier to choose correctly under real-world constraints. That’s where senior operators still matter. Not because they’ve “been around longer,” but because they’ve: seen second-order consequences survived decisions that looked right and went wrong anyway learned which constraints are real vs. theoretical A 22-year-old with AI can: * build faster * explore more * prototype endlessly But they often don’t yet know what matters when tradeoffs collide. And that’s not a knowledge problem. It’s a consequence problem. The real divide isn’t age It’s people who are willing to update their decision-making vs. people who aren’t. There are: 28-year-olds already locked into “this is how it’s done” 55-year-olds experimenting like crazy with AI every day The split isn’t generational. It’s behavioral.

The software era taught the market to price access. The Agentic Era will be built on something else. At ShareAI, we believe AI is built by many, so its upside should be shared by many. We are building the economic layer for AI: a new foundation where the people who power AI, build with AI, and create with AI can all share in the value generated by every inference. In this video, our founder explains why old software-era monetization breaks in the Agentic Era, and why a new model has to take its place. That is the shift behind ShareAI: from static software pricing to an economic layer built for real AI usage, builder monetization, and shared upside. Soon, we will reveal what comes next. AI for humankind. The economic layer for AI.

We live in weird times. You'll see on one side people sharing a small stack of Mac minis because they're building an "AI army". Then some're leaning into their quirks because they believe it makes them "AI bulletproof" and are inviting others to do the same. The thing is,  both are missing the same thing. AI can only amplify what already exists. It can't create something from nothing. So, if you're an expert in your field, AI helps you do more, work faster, and create things you've always wanted but didn't know how to. But if you're not an expert, no tool, prompt, or AI agent changes that. It just makes you feel like it does. That's what my LinkedIn feed is all about these days. A few experts showcasing their expertise with AI and people using AI to fake being something they're not (and we're all seeing through it).

We trust AI to approve loans, screen candidates, and flag diseases; but do we actually understand *how* it makes those calls? 🤖That's the question that pulled me deep into Explainable AI (XAI) this week. Here's what struck me most: accuracy alone is no longer enough. A model that predicts correctly but can't explain itself is a liability in high-stakes domains; medicine, finance, law, hiring. We need AI systems that are not just smart, but *legible*. Some of the key ideas I explored: 🔍 SHAP & LIME — post-hoc techniques that reveal which features actually drove a prediction, even inside complex black-box models. ↩️ Counterfactual Explanations — instead of just saying "rejected," the model tells you: "If your debt-to-income ratio were 10% lower, you'd be approved." That's actionable. ⚖️ The core trade-off — simpler models are easier to explain but often less powerful. XAI is about navigating that tension thoughtfully, not ignoring it. My biggest takeaway? Explainability isn't a feature you bolt on at the end. It has to be a design principle from day one; in your data choices, model selection, and governance structures. As AI moves deeper into decisions that affect real lives, the question shifts from "Can the model do it?" to "Can a human understand, verify, and contest it?" If AI is the engine, explainability is the dashboard. 🚗 Researching this genuinely changed how I think about building models — it's not just about getting the prediction right, it's about being able to stand behind it. #ExplainableAI #MachineLearning #ArtificialIntelligence #ResponsibleAI #XAI #AIEthics

AI is giving us incredible speed… and new ways to get lost. It’s now possible to build in a day what used to take a week. That’s exciting. But it’s also easy to feel like you’re making progress while actually heading toward a dead end — especially if you don’t fully understand what’s happening under the hood. Without strong foundations, speed becomes risky: You skip step-by-step thinking You end up with messy, hard-to-maintain code Even AI struggles with overly complex systems The people making the biggest leaps with AI aren’t skipping learning — they already have it. AI just amplifies their skills. AI is the engine 🚀 Foundations are the steering wheel Without both, you’re just crashing faster.

Day 1 of an AI partnership is the worst it will ever be. Your AI knows nothing about you. Your preferences, your communication style, your business context - all blank. Every interaction starts from zero. It feels clunky. Generic. Barely useful. Day 100 is a different universe. By Day 100, the system knows your writing voice. It knows which clients need careful language and which prefer blunt feedback. It knows your team's strengths. It anticipates the question behind your question. This is the compound effect that nobody talks about. AI partnerships don't improve linearly. They improve exponentially once you cross the context threshold. The AI that frustrates you in week one becomes indispensable by month three. But only if you build for persistence. If your AI resets every session, you never get past Day 1. You're trapped in an infinite first date, explaining yourself over and over to something that should already know you. The gap between companies that get value from AI and companies that don't isn't the model. It's whether they invested in making Day 100 possible. New on The Neural Feed: https://lnkd.in/euh9NV7T How long did it take before AI felt genuinely useful in your workflow? Days? Weeks? Still waiting?