Adriana Munoz Vergara - Chief Growth Officer at Domu Artificial Intelligence and Growth Specialist.
Soon, AI-native companies will not sell tools to services businesses; they will be the services business. According to Sequoia, the next trillion-dollar company will be a “software company masquerading as a services firm,” with the potential prize enormous, since “for every dollar spent on software, six are spent on services.”
I agree with this conclusion. Software fought over IT budgets. AI-native services go after labor budgets, which Bessemer sizes at about 13% of U.S. GDP versus roughly 1% for IT. The playbook is common knowledge: Sell the outcome, price around the work, start where it is already outsourced.
But after two years of agreeing on this, most founders still pick their sector by the size of the labor budget. That is a fine way to size a market, but it tells you nothing about who keeps the business.
The Two Questions I Ask Instead
Can you tell, quickly and cheaply, whether the work was done right? And does that verdict come back to you or live in the customer’s system?
The first framing question is inspired by Jason Wei, who wrote that the ease of training AI to solve a task is proportional to how verifiable the task is. He was referring to models, but I believe the same logic governs whether a services company compounds.
As a whole, work can be sorted into three tiers:
1. At the top, verification is nearly free because it is a by-product of doing the job. Did the customer pay? The system knows in days, in dollars. The work grades itself.
2. In the middle, verifying the work requires nearly as much labor as doing it. A senior lawyer reviewing a junior’s markup rereads the whole agreement, which is why “AI drafts it, a human checks it” can save less than a demo may suggest.
3. At the bottom, no clean verdict arrives. A strategy recommendation may take weeks to produce and years to judge, all because you never observe the counterfactual.
The second question concerning the verdict is the one people skip, yet I find that it decides more. If the outcome lives in the customer’s system, doing more work teaches you nothing. You can be the most accurate vendor in a category and still fly half blind, seeing only the inputs the customer hands you.
Why The Hardest Corner Is Usually The Best One
Here the standard advice inverts. Rank sectors by labor spend and you land on consulting, insurance, legal, financial services. The biggest pools are also where the verdict is slowest and the incumbent’s name matters most.
The categories that score highest on verification and ownership are the ones founders tend to avoid because they look painful. Regulated work is the clearest example: slow to enter, licensed, liable, governed by rules like the Consumer Financial Protection Bureau’s (CFPB’s) Regulation F, which sets federal standards for how and when you can even contact a consumer about a debt.
Yet, that is what I think makes it good ground. Regulation forces you to be accountable for the result, not for one step of the workflow. To be accountable, you have to own the whole interaction. And when you own the whole interaction, the verdict has nowhere else to go. It closes inside the company that made the decisions.
The Value Of A Closed Loop
I have watched this from inside a regulated category, and the lesson that transferred was structural. Every interaction is an experiment in timing, channel and offer, and the result lands back in the system that chose them. You learn not what was said, but what happened afterward: what was paid, converted, renewed, churned, delivered as promised.
This is why the tool-plus-incumbent combination tends to underperform. A great AI tool inherits the customer’s org chart, so it can make one silo faster but cannot force three silos into one loop.
MIT’s researchers studied 300 public AI deployments and found roughly 95% of enterprise generative AI pilots produced no measurable impact on profit and loss, yet companies working with AI specialists outperformed those building solutions internally by about 3-to-1. Specialists win because they own an outcome end to end. Everyone else owns activity.
Pricing is the honest tell. When a provider can charge for a resolved outcome rather than for seats, hours or usage, it signals that it can verify the result and control enough of the chain to be responsible for it.
The Counterargument Worth Taking Seriously
Some services survive automation not because the work is hard, but because the client is buying reputation as much as output. They want the right to say a credible name stood behind the result, and that if it goes wrong, the judgment was not theirs alone. A better model makes the work cheaper, but it does not confer a Big Four signature.
Slow verification therefore protects incumbents twice. It gives the challenger slower feedback, and it makes clients care more about the brand the incumbent already has.
So the real question is which side can develop what it lacks first? Can the challenger build trust around a closed loop, or can the incumbent close the loop before it loses the work?
What You Should Take From This
Before defining the market, get clear on what success actually looks like. Write down the moment that proves the work is done, using language the customer’s finance team would recognize. If you cannot explain it in one sentence, you may be selling activity rather than an outcome.
Then check where that verdict lands. If it lands in the customer’s system, your road map should be about getting closer to it, even when that means unglamorous, operationally heavy work.
And measure pull: How many customers ask you to run an outcome you have not sold them yet? Road maps are free. Pull is the only proof the category is real.
In the end, the size of the labor budget is not the question. The question is whether doing the work teaches you anything, and whether that lesson stays with you.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?