Why Becoming AI Native Is Harder Than It Looks

Ask a finance firm what it has actually done with AI, and the most common answer is that everyone got a chatbot. It’s worth asking the follow-up: are the deals any better? Faster memos and quicker screens are easy to point at. Better deals, a clearer funnel, a higher pass rate, those are the numbers that matter, and they’re usually met with a pause.

That pause is the whole subject of this piece. Becoming AI native is much harder than it looks, and the reason is counterintuitive: the first few weeks feel like real progress, and that feeling is precisely the trap.

L.E.K. put a number on the gap recently. It surveyed 100 US buyout professionals, and only 21% rate their own AI expertise as advanced. The same group reports an average 28% productivity improvement from AI already. Honestly, I’d be skeptical of even the 21%, because advanced is a relative word and hard to quantify, and from what we see day to day the real figure is much lower. Either way, the gains are real and measurable, and the overwhelming majority of the people delivering them don’t yet know what they’re doing. The space between what the tools give you on day one and what competence actually takes is where nearly everyone is standing right now.

Day one is free, and that is the problem

A chatbot is useful the first time you touch it. You type a question, something competent comes back, and the sense of mastery arrives immediately and at no cost. Nothing else in software has ever worked like that. Excel was hostile for a month. A CRM rollout took a quarter and a consultant. You knew you were a beginner because the tool reminded you constantly.

A modern AI assistant never reminds you. It stays helpful and articulate right up to the point where the work needs real craft, and then it produces something that still looks fine. The output degrades far more slowly than your competence does, so there is no signal marking the edge of your ability, and most people cruise straight past it. That is why the day-one feeling of competence is so misleading: it is the same whether you are one hour in or one year in.

The blunt version, and I have said this to clients: you have to crawl before you sprint, and most firms are trying to sprint before they can crawl.

The same ceiling, in a nicer font

The trap has a spectacular version doing the rounds at the moment: the firm that replaces its expensive CRM with a vibe-coded one built over a few weeks, taking a six-figure annual bill to zero. The usual objection is that it will not survive the person who built it leaving, which is fair, and also one step short of the real problem.

When you rebuild the thing you already have, your requirements are a description of the system you are replacing, because that is the only version of the job you have ever seen. You confine yourself to everything you previously knew. So you arrive at exactly the same ceiling. Same fields, same reports, same blind spots, and what has changed is the font, where the buttons sit, and the fact that you own it now.

It is the most convincing form of false progress there is, because you did build something, it does work, and it did take real money off the bill. Every signal says you have moved. You are standing where you started. The bar for what software can be has just risen, on capability, quality and personalization, and rebuilding yesterday’s tool without yesterday’s license fee does not clear it. This is a close cousin of the build-versus-buy trap we have written about before: the instinct to replace overpriced software is right, but building your own version is usually a trap for exactly this reason.

The question nobody in that argument asks is what the system should do if it were designed today, by someone who understands both the technology and the work. That version does not get built in three weeks, because most of the work sits underneath the interface, in the structured record a system needs before it can do anything the old one could not. One lower-middle-market PE firm we worked with could not answer “why did we pass on this?” without half a day of folder archaeology; turning five years of scattered deals into one queryable, cited foundation was slow precisely because that record is the part that matters.

The system underneath you

There is a second failure mode, and it gets far less airtime. You try the tools, nothing miraculous happens, and you conclude the whole category is oversold. Fair enough, given what you were promised. The trouble is the promise was pointing at the wrong thing.

Everybody wants the analyst in a box: the associate that builds the flashy model and drafts the whole memo in one pass. Right now most of the real value sits somewhere much less exciting, in the systems that keep the work running, and we badly underrate how much time they eat and how much they are worth. A solo advisor told me recently that he keeps his deal trackers partly in a spreadsheet and mostly in his head. His question put it better than I could: how often do I update those trackers in a mad week? Never. The task takes thirty minutes, thirty minutes does not exist in a mad week, and the tracker stops being true.

The same goes for keeping contact lists current, tracking a referral network, and the obvious one, keeping the CRM up to date. A boutique M&A advisor we work with had exactly this problem: the pipeline lived in a spreadsheet updated by hand, the relationships lived in people’s heads, and when a banker left the knowledge went with them. We connected the firm’s email and calls into DealSage so the CRM populates itself, capturing every interaction and outcome against the contact automatically. The point worth noticing is what we did not do: we did not hand the judgment to a model and ask it to run the relationships. We took the admin off the humans and left the deal-making with them, which is where the parts of the job that don’t scale were always going to stay.

Underneath all of it sits the information itself. You come across valuable pieces of it all day, on calls, in emails, in passing, and mostly you do nothing with them. Capturing that, organizing it, storing it, linking it, doing the backend work that makes it start to compound, this is what AI is genuinely good at now. You are building a system underneath yourself, and that is where the real strength lies.

That, for me, is what becoming AI native means. Not the analyst in a box, and not the weekend rebuild of software you already had. A system built underneath you, one unglamorous piece at a time. It takes a lot longer than a weekend, and it compounds for years.