|
Getting your Trinity Audio player ready...
|
Three conversations boutique founders keep having about AI and the one they’re avoiding.
Two years into the AI era, many boutique professional services founders I talk to are stuck. The initiative isn’t dead, but it isn’t really moving either. There’s activity. Meetings, demos, vendor calls, but very little has actually shipped. No new economics on the P&L. No real change in how the firm operates. No buyer would look at the business and call it AI-native.
When I ask why, I usually get one of three answers. We’re still figuring out the technical architecture. We hired a vendor and it didn’t work out. We’re evaluating which tools to standardize on.
All three answers sound reasonable. All three feel like progress. None of them are.
The honest diagnosis is that the founder is having the wrong conversation and using the wrong conversation to avoid the only one that matters.
We’ve seen this movie before
I’ve spent twenty-five years in product management. Long before AI was the thing every founder was trying to figure out, I was watching new products fail at staggering rates. The standard estimates run from 70% to 90%, depending on the study, but the direction is consistent. What’s even more consistent is the reason products fail. They fail because teams build before they understand the problem. They fall in love with a solution. They obsess over what’s technically possible instead of what’s actually needed. They get stuck choosing a vendor or a platform when they haven’t yet decided what they’re trying to do for whom.
Uri Levine, the co-founder of Waze, summarized the lesson product managers eventually learn in one sentence:
“Fall in love with the problem, not the solution.”
That’s the single most important sentence in product management, and most teams discover it the hard way. By spending time and money on a product nobody wanted.
We are not doing this for the first time with AI. The patterns are familiar. The failure rate is familiar. The remedy is familiar. It’s the same one product managers have been preaching for decades: figure out the problem before anything else.
That’s the lens to keep in mind as we walk through the three places founders get stuck. Each one is a familiar product management failure.
It’s not a technical problem
The first failure mode is treating AI like an IT project. Founders bring in a CTO or a “head of AI,” scope the architecture, debate whether to build or buy, argue about model selection, draw integration diagrams. It looks rigorous. It feels like the way grown-up companies do things.
In product management, this failure has a name. It’s called solution before problem. The team commits to a technology, a feature, or an architecture before anyone has clearly defined what problem the product is supposed to solve, for whom, and under what constraints. The output is always the same. A polished thing that doesn’t work, because the question it answers wasn’t the question that needed asking.
AI in a boutique professional services firm is NOT a technology rollout. It’s a use-case discovery problem, and discovery is something product managers know how to do. The firms making real progress aren’t the ones with the best stack. They’re the ones who picked a real problem, something specific enough to measure and large enough to matter, and started solving it.
There’s a second reason the technical-first frame fails. The technology doesn’t sit still. I tell members all the time: if AI can’t do something today, wait until Tuesday and it probably can. The capability frontier moves weekly. Founders who over-engineer around today’s limitations are designing for a world that won’t exist by the time the design is done. Pick a use case, start solving it with what’s available, and let the technology improve underneath you. It will.
It’s not a vendor problem
The second failure mode is hiring outside the firm to figure out AI for you. Founders sign up consulting partners, AI implementation shops, the latest tools, anyone who looks like they might know more about this than the founder does. The thinking is, I’ll outsource the figuring-out, then bring it in-house when it’s working.
It almost never works, and the reason isn’t that the vendors are bad. The failure here is more fundamental than vendor quality or selection. In product management terms: you can outsource implementation. You cannot outsource problem definition.
The person closest to the customer, the economics, and the strategy of the firm has to define the problem. That’s not a vendor, that’s the founder. No outside party can make that call. They don’t know your clients, your pricing, your delivery model, operating model or your economics well enough to decide which problems are worth solving. They can build it but they can’t decide what the it should be.
Matt McElyea’s piece on what happened at HumCap when their AI vendor shut down overnight is required reading on the tactical side. Vendor risk is real and needs to be designed for.
Problem definition is founder work. It can’t be delegated, and it can’t be bought.
It’s not a tool problem
The third failure mode is the most seductive because it feels the most concrete. ChatGPT or Claude or Gemini. n8n or Zapier or Make. Cursor or Windsurf or Copilot. Founders convene meetings, run bake-offs, build comparison spreadsheets, debate the merits of different vector databases. It looks like a decision is being made.
A decision isn’t being made. A decision is being delayed.
Anyone who lived through the marketing technology wars of the 2010s has seen this exact movie. The same firms that spent two years deliberating between Salesforce, HubSpot, and Marketo and ended up no further ahead than the firms that picked one and went for it. This is the same playbook today but with AI tools. In 2026, the leading tools are all good enough. The differences between them matter at the margins, and the margins don’t matter until you’re actually using the tools to solve a problem.
Most founders standing in the tool aisle aren’t there because they’re picky. They’re there because the tool conversation is more comfortable than the use-case conversation. Tool selection has clear inputs, clear vendors, clear pricing. Use-case selection requires the founder to make a judgment call about where the firm should be heading. That’s uncomfortable.
Progress over perfection. Pick anything serviceable. Just go. The tool stack will sort itself out once you’re moving.
What founders are actually avoiding
Strip away the technical conversations, the vendor conversations, and the tool conversations, and what’s left is the conversation founders don’t want to have. What problem in this firm is worth using AI to solve? Who is this for? What outcome am I committing to, and what does success look like?
These are the first questions a product manager asks at the start of any project. They’re so foundational they feel basic. They’re not. They’re the hardest part of the work, because they can’t be hedged. You can’t bring in a vendor to make the call. You can’t blame a tool if the call is wrong. There’s no architecture diagram that lets you defer the decision another quarter. It’s just you, picking a thing that matters, naming what good looks like, and going.
The fast eat the big
There’s one other thing worth saying about this moment. Most of the time in business, the big eat the small. In transitions like this one, the fast eat the big.
Boutique firms have an advantage right now that they don’t usually have. Big firms move slowly. They have committees, governance reviews, procurement processes, change management programs. They’re optimized for the world that existed last quarter. Boutiques can decide on Monday and ship by Friday. Most don’t, because they’re imitating the deliberation of larger firms instead of using the velocity that’s actually their edge.
Product managers know this instinct as a release cadence problem. Teams that ship every two weeks beat teams that ship every six months. Not because they’re smarter, but because they get more learning cycles per quarter. Boutique firms in the AI era are in the same position. The ones that win the category won’t be the ones that deliberate better. They’ll be the ones that ship faster, learn faster, and adjust in flight while the big firms are still in the framing meeting.
A test for whether you’re stuck
If your last three AI conversations inside your firm have been about technology, vendors, or tools, you’re stuck. Not because those conversations don’t matter, they do. Eventually, but they’re all downstream of a decision you haven’t made yet.
Try this instead. Pick one problem in the firm. Name what success looks like in concrete terms: a number, a margin, a cycle time, a client outcome. Commit to a date. Tell your team. Then start. The tools, the vendors, and the architecture will reveal themselves quickly once you’re actually trying to do something.
After twenty-five years of watching this pattern, the difference between firms that get unstuck and firms that don’t is almost always the same thing. It isn’t talent, capital, or technology. It’s whether the founder is willing to act like a product manager about their own AI initiative. A founder willing to fall in love with the problem, name the outcome, and ship something before they think they’re ready.
The founders who do will be the ones their buyers describe, two years from now, as the AI-native firms in the category.
Just go.