AI-native consulting is here. It looks the same in every domain.

Legal went first. Contract review and demand letters are now sold per document, by firms built around AI from day one.

The same thing is now happening in four more domains that matter to anyone running a services business: accounting, customer service, go-to-market, and IT.

I looked at the firms leading it. They don't sell AI. They take well-defined jobs off the client's hands, run them, and charge for the finished result. Different industries, different founders, different scale — the same pattern, eight characteristics, every time.

What is emerging, domain by domain

In every one of these domains the same players are at the table. The advisors, who tell you what to do. The platforms, whose software now does more of the work itself, with AI agents inside the licence. The integrators and outsourcers in between, who implement and run it — that is where most services revenue sits today. And a new kind of firm coming for that middle ground.

Accounting

Two firms rebuilt the accountancy itself rather than selling software to accountants. Both were founded by people who had built accounting software and concluded that the software was the small part of an accountancy's cost — the people were the big part — so the leverage sat in owning the firm, not the tool.

Both sell a fixed price per client. No billable hour. One handles accounting, tax and payroll under one roof — one relationship for everything a business owner needs — with licensed experts checking every number; it bought a payroll firm to complete the set and plugged itself into a business bank so clients run their books from the account they already use. The other went narrower still: bookkeeping for one type of client, with no fixed accountant per file. The system remembers every conversation, so whichever senior is free picks up the file already briefed. Three numbers are never traded against each other: client satisfaction, staff satisfaction, files per accountant. The firm chooses its clients, and the human's job is exceptions and the signature.

Customer service

Customer service is the domain furthest along, because the middle ground was already occupied. Contact-centre outsourcers have taken the job off the client's hands for decades, and have long priced on volume — per agent hour, per contact — rarely on the result. Now the AI agent resolves the routine request end to end, the human agent handles the exceptions and the emotional moments, and the price moves to per resolution — or to a share of the result: satisfaction held, cost per contact down, churn avoided.

The large outsourcers are rebuilding their delivery around this, and the integrators are buying them to get there faster. Alongside them, new firms sell one thing only — a customer request, resolved — with no seat and no licence attached. The platform vendors have followed: they bill per resolved conversation too. Everyone in the chain now sells per result. The question for the CX integrator in the middle is what it sells that the outsourcer and the platform don't.

Growth and go-to-market

On the intelligence side, the most recognisable consulting job of all — commercial due diligence for private-equity investors — is now delivered in days instead of weeks by a new consulting firm built for it. AI agents do the interviews, gather the data and draft the analysis. A senior consultant frames the questions and signs off on every finding. The client can buy the full study or only the three questions that decide the deal. Every claim in the report links back to the interview or dataset behind it. The same firm sells voice-of-customer and market-entry studies to corporates the same way. The client pays for the answer, not for the weeks.

On the demand side, new agencies take the whole front end off the client's hands: from the target list to the booked meeting. Agents research each account, write and run the outreach, and keep everything they learn about the client's market from one campaign to the next. A senior person designs the play and takes the conversation once a buyer answers. The agency is paid per meeting booked or per pipeline created, not per campaign hour.

IT

In IT the same players are there, and the platforms have moved fastest: they resolve tickets themselves now. The advisors sell maturity assessments. Implementation consultancies have started publishing fixed-fee sprints and 90-day builds with a written guarantee — a real step towards outcomes, but still a build handed over. And the middle has just been taken — by acquisition, not by founding.

A managed-service provider that has run IT for financial firms for decades was bought by an AI-engineering group, and spent a year rebuilding how it works. Agents now gather the diagnostics, correlate signals across systems and propose the resolution; the engineers stay in the loop for judgement and escalation. One AI layer sits under everything the firm delivers — IT, security, cloud, compliance — and the agents it delivers are certified against its own trust framework. It reports the results every month: satisfaction surveyed on every closed ticket, response times against SLA, and ticket volumes down sharply in the first year once root causes are fixed. It calls itself the first AI-native MSP, and it has a point.

What they have in common: eight characteristics

Everybody in these markets is moving towards outcomes. An outcome, in the sense these firms use it, is a finished unit of work — not a promise about your P&L. The new firms take responsibility for that finished job: not for a recommendation, and not for a piece of software. Eight things follow from that.

1. One clear job — or a short catalogue of them — end to end. They don't sell AI for a domain. They pick a job the client already buys from someone, or a small set of clearly defined jobs, and own each one completely. This only works when the firm owns the whole job and the result can be counted; that is why they define it so sharply. The new price point then widens the market in one of two directions: downward, to clients the old fee excluded, or inward, to work the client used to keep in-house.

2. They sell a finished job, priced against what the client used to pay. Not hours, and not a share of your result — a unit they own and can count: the report, the file, the resolved conversation, the booked meeting. Nothing to dispute afterwards: no baseline to agree, no debate about whether the result came from the firm, the tool or the client's own team. The saving is the difference between the old invoice and the new one. The P&L effect is in the price, not in the contract.

3. The senior person is the frontline. "Led by senior consultants." "The accountant signs." "Engineers in the loop." The senior frames the question, handles the exceptions and signs the result. The agents do what the junior layers used to do — the gathering, the drafting, the first pass — and that layer thins or goes. What the client pays for is judgement.

4. They build their own intelligence layer — and it learns. Every firm rents the same AI models. What they keep is their own: the traceable evidence base, the client memory, the AI layer under every service. And it learns — every approval, every exception, every closed file makes the next one cheaper and more accurate. The second engagement is cheaper than the first, and the hundredth is cheaper still.

5. Trust is built into the offer, not claimed. Every claim traced to its source. A satisfaction score the firm refuses to trade for volume. Agents certified against a trust framework. Governance documents in the auditor's own format. When part of the work is done by a machine, trust is the product.

6. The first step is small, fast, and stands on its own. Three questions instead of the full study. An eight-week roadmap before the recurring service. Per resolved conversation, with nothing to install. Each is priced, delivered quickly, and useful even if the client stops there — and each leads naturally to the next step. The old firms asked for a large commitment before proving anything. The new firms prove it first.

7. They were built by insiders. A former investor. A former strategy consultant. Founders who had sold accounting software. Managed-services veterans paired with AI engineers. They knew the workflow and the price before they built the firm.

8. The homepage says what they are in one sentence — before it says what they do. "The AI-native consulting firm." "The first AI-native technology partner for finance." Some also say what they are not, including when you should not hire them.

And the established firms?

The established firms are not standing still. Across every domain they are making one of four bets.

Some are rebuilding the run by embedding AI in their own delivery. The strategy house whose internal assistant cuts weeks off the analysis — while the fee stays where it was. The accountancy network that puts AI under its own delivery, so the same team handles more files. The outsourcer whose agents resolve the routine contacts, still billed per agent. These firms are changing how the work gets done, and it shows in their margins. What it does not yet change is what the client pays for. Rebuilding the run and keeping the hour is efficiency. Rebuilding the run and repricing the unit is the AI-native move — and it is the step most have not yet taken.

Some are jumping into software. Services firms build a product, launch it as a separate company or brand, and sell services next to it. The services stay what they were; the product is a bet on becoming a software business — with software economics, software go-to-market, and usually software funding. It is a real move, and a bigger one than it looks. But it is the platform route, not the AI-native one. The AI-native firms build their own software too, and keep it: they run the job with it and sell the finished result. The side-by-side firm sells the software, and the client does the work.

Some make AI the subject rather than the model. The audit-and-advisory network that funds an AI company's growth round and gains an AI story with it. The specialist consultancy that sets up a digital practice beside its premium core. The accountancy that offers "custom applications with AI" through a partner, as one more line in the catalogue. The alliances with model vendors, announced with a very large number attached. In each case AI becomes something the firm sells or points to — a new practice, a new competence, a stake — while the way the firm itself works and charges stays as it was. Real capability, sometimes real revenue, and much of what is announced is old automation with a new name. AI-enabled, not AI-native.

The second and third bets share a trap. A practice that stands beside the firm's reputation inherits none of it. It competes with agencies and integrators at their prices, and the premium the core has spent decades defending stops at its door. The test is simple: can a partner in another office explain in one sentence what the new practice sells, and to whom? If not, it sits beside the firm, not on top of it.

And some are defending the ground they hold — by going premium where the new firms go cheap. The new firms sell a clear catalogue of jobs, pay-per-unit, at a fraction of the old price. That leaves room above them. Predictability: the client knows the cost, the scope and the date before signing, and the supplier carries the risk on all three. Budgets are annual, results are not, and procurement cannot approve a bill that swings with volume — buyers say they will pay a premium for that certainty, and the large firms already sell it. Compliance: in regulated environments the client pays for the firm that carries the liability, passes the audit and can stand in front of the regulator. Resolving the ticket is table stakes; being accountable for it is the premium. Breadth: the new firm owns a set of clearly defined workflows; the incumbent can own everything around them — the integration, the change, the exceptions, the hard cases that need senior judgement.

This holds on two conditions. The premium has to be packaged and priced as such, not the old portfolio with a new adjective. And it works where the result is genuinely hard to isolate; where it can simply be counted, the new firm's price wins.

None of these four is wrong. They are bets on where the firm wants to sit: how the work is done, whether to build software, what is sold, what is charged. The question that separates them from the new firms is the same in every case: who does the work, and who gets paid for the finished job?

Follow the money and the answer is visible. Strategy houses buying engineering firms. Integrators buying outsourcers. Outsourcers building their own platforms and moving to outcome pricing. A new kind of investor buying ordinary service firms and rebuilding them around agents. Everyone is moving to the same place: take the job off the client's hands, run it, and price the result. Some of these firms will fail. The pattern won't.

Notice the order

Every one of the new firms decided four things before it decided anything about technology. Which customers. Which job those customers already pay for. What the finished unit looks like. What it costs.

None of them invented a market. Every one of them took a segment that was already being served — expensively, slowly, by the hour — and built an offer fitted exactly to those customers and that job. Nothing broader.

Only then did they build the operating model: the agents, the platform, the human in the loop. And they built it to keep getting better — every file, every conversation, every ticket making the next one cheaper and more accurate.

Customers, job, unit, price. Then the machine. Then the machine improves.

That order is the whole difference between being AI-native and being AI-enabled.