> ## Content Index > Fetch the complete content index at: https://www.vysora.co/llms.txt > Use this file to discover other available public pages before exploring further. # We Sold More. Founders Needed Less. AI Changed the Math. - URL: https://www.vysora.co/we-sold-more-founders-needed-less-ai-changed-the-math/ - Published: 2026-10-19T14:08:11.000Z - Updated: 2026-10-19T14:08:11.000Z - Description: The founders who bought less often built better products. Why agencies sold them more, and how AI changes the math for both sides. - Author: Enes Lisovac - Tags: AI Downstream, Startups, Agencies, Founders, Product development The founders who bought less from our agency were often the ones whose products did better. Some came knowing exactly what they needed next: an assumption validated, a prototype to test a specific solution, a designer for one flow, an integration feasibility check, an engineer for scoped work. They questioned every extra phase, role, and week. They bought what they needed and stopped there. Others came with an idea. Some had done research or had early customer interest. Others wanted their own version of a product that already existed. What they wanted from us was help deciding what to build, how to build it, and what should come next. For that second group, we often sold the full package of our services: discovery, product strategy, branding, design, engineering, MVP delivery... Bigger engagements. In my experience, fewer of those products found traction. In ["We Sold Projects. Founders Needed a Loop. AI Changed Who Owns It."](https://www.vysora.co/we-sold-projects-founders-needed-a-loop-ai-changed-who-owns-it), I wrote about the mismatch between selling projects and building products that have to keep learning. That mismatch was visible for years. What kept it in place wasn't bad advice or bad people. It was economics, and the founders least able to see it often paid the most. AI is now changing those economics from both sides. It gives founders a real alternative. And for agencies, it changes more than the cost of effort. It challenges both what they sell and how they sell it. Effort priced by the hour gets cheap, founders can build more of the first version themselves, and phased proposals become much easier for founders to question. --- ## The Next Phase Was Always Ready We’d deliver the package phase by phase until the MVP was ready, and before launch we’d already be selling the next phase: production support, growth services, new features... Each proposal made sense on its own. The product would need hosting and monitoring. Users would need better onboarding. Investors would want a roadmap. Then reality hit: too few users, too little revenue, a fundraise that didn't close, runway running out. Many founders couldn't afford the next phase. We had delivered everything we agreed to. By the contract, the engagement was a success. For the product, the real test was only beginning. Some founders offered sweat equity: a stake in the company in exchange for continued work. The agency would accept when it believed in the founder and the product, at least to some extent, and when the founder could show some proof that a fundraise was coming. On paper, it aligned incentives. The agency would share the upside, so it would care about outcomes rather than hours. While the initial terms held, usually a deadline for the founder to raise, everything was fine. Then the fundraise took longer than planned. Equity on paper doesn't pay salaries. Every month, the agency carried a team on a project that was now losing it money, while it still needed margins to run the business. The pressure moved to the founder. If the founder couldn't raise enough, or not in time, the agency started moving people to paying projects. Gradually at first. The team shrank to what production needed to keep running. Then that went too. Sometimes the round closed and it worked. Too often, it only prolonged the struggle. Most of these projects failed for both sides: the founder's product didn't make it, and the agency lost the money and the margin it had put in. I'm not claiming agencies alone caused those startups to fail. Most startups fail, with or without an agency. But the business model contributed. We packaged uncertain product development into phases and sold each phase as a project with a budget and a deadline. Every phase consumed runway and deepened the founder's dependency on us. So when users challenged the assumptions, there was less money and less room to change direction. --- ## Nobody Has to Be Doing Anything Wrong Advising founders to buy less was one thing. Figuring out how to pay more than a hundred agency salaries was another. I've been on every side of that conversation. I started as a software engineer. In architect roles, I helped sales teams design and write proposals, and I was pulled further into the sales process over time. Later, I ran my own agency and sold these myself. Then, in VP and C-level roles at other agencies, I shaped how they were packaged and sold. In every one of those roles, selling was a balance: how much to advise, how much to propose, when to tell a founder not to spend the money yet. I never pushed founders toward the biggest deal I could close. I knew what every *"you don't need this yet"* cost us. And it was rarely a clean call. Some founders were completely sure of their idea. They had the money, or said they could raise more, and wanted far more built than we thought they needed. Sometimes I believed the product wasn't promising, or the scope was far too big. But I couldn't know it would fail. Products succeed for reasons no plan captures, and product sense is more than spotting a bad plan. It wasn't our place to tell a founder no, or to talk them out of their own company. So we advised, and then we built what they asked for. I never made peace with that balance, and over time I started questioning the whole approach. A founder comes in with an uncertain product idea. The agency knows how to turn it into a proposal: discovery, design, engineering, delivery. Roles, weeks, hours, rates, margins. Sometimes discovery is genuinely valuable. Good research, a working prototype, and clearer assumptions prevent expensive mistakes. But reducing uncertainty isn't eliminating it. Real learning continues when people start using the product, and by then the next phase is usually already sold. Behind every proposal sits an organization. People on payroll. Teams that need their next project before the current one ends. Utilization to maintain. Margins to hit. A revenue forecast that assumes the pipeline converts... An agency sells expertise, a ready-made team that can scale up or down, and the feeling of certainty that comes with a full plan. It packages them as a fixed-price project, a retainer, or time and materials. Whatever the package, it gets paid through effort: people, weeks, hours. Inside it, everyone acts reasonably: - Sales wants to win the deal, and a bigger deal covers more of the forecast - Engineers want to build something reliable, and sometimes imagine a level of sophistication the founder doesn't need yet - Delivery protects scope, timelines, and budget, which in practice means protecting the plan from evidence that arrives after it - Leadership needs every team booked next quarter None of them has to be doing anything wrong. Collectively, they can commit much of a founder's runway to problems the business hasn't yet proven it needs solved. I kept arguing for something smaller. Simpler architecture. Narrower scope. Cheaper hosting. Fewer commitments. More room to learn before investing further. But I wasn't just arguing with a salesperson, an engineer, or an agency owner. I was challenging the economics holding the whole organization together. You can't ask an agency to consistently sell less without answering how it survives doing that. And I didn't have that answer. --- ## Who Was Least Equipped to Say No The first group of founders faced the same uncertainty as everyone else. What they had was better product judgment, usually built on domain knowledge and real contact with customers. That shaped what they asked for, what they refused, and how their products turned out. Buying less was a side effect of knowing what they needed to learn - or build - next. They outsourced the work, not the learning. The second group is where the model did the most damage. The founders who most needed help deciding what and how to build were often the least equipped to judge our proposals. Consider what judging a proposal requires. Is discovery needed, and how much of it? Does this MVP need this architecture, this design system, this infrastructure? Is a twelve-week plan reasonable, or would three weeks answer the question that actually matters? Every one of those calls takes the product and technical judgment the founder came to buy. So the founder relies on the agency to define the work, estimate it, and decide whether it's necessary. The party selling the effort also decides how much effort is needed. Most professional services work this way, from renovations to legal advice. But in early product development, most of the scope is a guess. The gap between what's proposed and what's needed can decide whether the founder gets one more experiment or none. A full package also looks safer. It answers every question upfront: product strategy, design, engineering, launch. For a founder who doesn’t yet know what they don’t know, a complete plan reads as reduced risk. In practice, it moves the risk. The founder stops worrying about not knowing what to build and starts carrying the cost of having built it. --- ## What AI Changes for Founders For the first group, AI amplifies an advantage they already had. They can build, test, and change more on their own, with even less need for agency services. The judgment that made them buy less now lets them build more themselves. For the second group, AI changes something more important: the cost of learning by doing. Product judgment comes from experience, and experience used to come from expensive attempts. Most founders could afford one serious attempt, and many spent it on a full package. Now a founder without product experience can put a working version in front of real users, get it wrong, change direction, and build judgment along the way, at a cost that leaves room to try again. AI also narrows the asymmetry. Before buying discovery, a founder can build a rough prototype and see whether anyone cares. Before signing a proposal, they can ask AI to take it apart: which phases matter at this stage, what the architecture is solving for, what could wait. For a founder who couldn't judge a proposal at all, that's real help, and it's new. But it isn't a neutral advisor. As I wrote in the previous article, it tends to agree with the hypothesis you bring to it, and it sounds as confident when it's wrong as when it's right. Ask whether a twelve-week plan is reasonable, and it will find reasons to say yes. Ask what to cut, and it will find things to cut. It can make a founder a better buyer. It can't be the judge. And there's the other outcome. A founder can use AI to turn untested assumptions into a polished product nobody wants, faster and cheaper than any agency could. AI makes the full package affordable without the agency. But the price of the package was never the real problem. Building before learning was. Even then, there's runway left to try again - if they actually learn. --- ## What AI Changes for Agencies Most agencies start in the same place: put AI in engineers' hands. Implementation gets faster. On fixed-price work, margins improve. On time-and-materials work, the hours shrink. That looks like progress, and for a quarter or two it is. Then the arithmetic catches up. AI reduces billable hours, not payroll. If revenue is *hours x rate* and AI cuts the hours, the same payroll needs more hours sold to stay covered. More phases. More features. Longer roadmaps. The pressure to sell more rises at exactly the moment founders need less. Effort got cheap. An organization built to sell effort now needs more of it to sell. Fixed prices only delay the problem. An agency can quietly keep the savings while founders keep paying for a two-week build that now takes two days. But founders can now see what a first version costs with AI, and so can the next agency pitching them. The price of "build me an MVP" is falling whether the agency lowers it or not. The quick answers are fewer people, lower prices, or a new label on the same offer. All three keep the same model. Fixing it is harder than it looks, because the incentive to sell more doesn't live in one place. It sits in the revenue model, proposal templates, contracts, sales compensation, utilization targets, roles built around handoffs between phases, and the way success gets reported: on time, on budget, as specified. Change one, and the others pull it back: - An agency that starts selling outcomes but keeps utilization targets drifts back to selling hours - One that drops fixed scope but still pays sales on deal size keeps proposing bigger deals - One that puts AI in engineers' hands but keeps the same proposals delivers the same phases faster, to founders who needed fewer of them - One that takes equity, as we did, ties its income to a few uncertain bets while salaries are due every month Underneath the structure sits the culture, and it's the slowest part to change. Agencies celebrate what keeps them running: the big deal announced to the whole company, the account that doubled, the team fully booked through next quarter, the project delivered on time and on budget... Bonuses follow revenue, utilization, and upsells. Promotions often follow the size of the team or the account someone manages. Nobody gets a bonus for the phase a founder didn't buy, or for telling a founder to come back later. Until that changes, every new model gets measured by the old scoreboard. The redesign has to answer questions the old model never had to ask. *What does the agency earn from when a founder needs less?* *How does it get paid for judgment instead of effort?* *What happens to roles that existed to coordinate handoffs AI no longer needs? How do you forecast revenue when engagements aren't sized upfront? How do you prove value to a founder who can build the first version alone? Which work do you stop charging for as AI takes it over?...* There's no single answer that fits every agency. Size, domain, clients, the people on payroll... all change it. Working through it with agencies is most of what I do now, and adopting the tools is consistently the easy part. The hard question isn't how many engineering hours AI saves. It’s whether you’ve built an agency that can afford to keep telling founders: "You don't need us for this. You can do it yourself." "You don't need us yet. Come back when you have paying customers." "Don't buy the next phase. Learn from this one first." --- ## The Learning Was Never for Sale The first group's advantage was never how little they bought. It was knowing what was worth learning next, and keeping that learning close while others did the work. For years, the founders who most needed that judgment had to buy help in its most expensive form: a full team and a full plan, sized before any user touched the product. AI gives them a cheaper way to build that judgment themselves: small attempts, real users, decisions they own. For agencies, AI removes the cover. When effort was scarce, a bigger proposal could pass for thoroughness. When effort is cheap, it's much easier to tell the two apart. Effort was only ever how agencies got paid. What founders valued was the judgment, expertise, and accountability behind it, and that's the part still worth paying for. I'd questioned this model for years without seeing a practical alternative. That contradiction was part of why I left the agency world. AI was the other part. First, it made the old model impossible to ignore. Then it made an alternative practical, for founders and for the agencies willing to rebuild around it. AI changes what founders need to buy. Not what they need to know. See AI Clearly. Think Beyond the Hype. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime.