This series has been about what agents do to enterprises. But there’s a whole industry whose entire business is selling that transformation — and whose own model happens to be the most exposed of all: consulting and IT services. If agents commoditize software priced by the seat, what do they do to advice priced by the hour? Who advises the disrupted, when the advisors bill by the person?
Services is the purest labour business there is. Strip away the brand and the frameworks and the model is people, sold by time. That makes it the sharpest possible test of the whole thread: the billable hour is the seat, and the delivery work — the models, the decks, the code, the tickets — is exactly the veneer that agents absorb. And yet the same firms are riding the biggest demand wave of their careers, because everyone is paying them to explain and install this shift. Both things are true at once — and which one wins depends entirely on the tier.
The strategy tier: winning — for now
The firms that sell judgment are, so far, the clear winners. AI-advisory demand has been enormous — a leading strategy firm reported roughly a quarter of its revenue from AI work, with total revenue growing and, tellingly, headcount growing too — but tilted hard toward AI engineers and data scientists rather than the traditional analyst intake. They’re the ones authoring the “agents are moving from pilots to core operations” manifestos, and charging to help act on them.
Why does the top of the market hold? Because what it actually sells — judgment, trust, access to the boardroom — doesn’t commoditize on a schedule. That’s the durable corner of the map. But even here, the base is eroding: the junior work that used to justify thousands of billable hours (build the model, do the research, make the deck) is exactly what an agent now does in an afternoon. The visible signal is unmistakable — top firms have frozen graduate starting salaries for multiple years running and quietly cut graduate intake, “extracting more value from fewer juniors.”
The professional-services tier: the barbell
The large professional-services and integration firms live in a genuine barbell. On one arm, they are booking AI transformation at a scale that dwarfs anything before — billions in new agentic-AI bookings, internal AI platforms with tens of thousands of seats, armies of consultants working alongside armies of AI agents. On the other arm, they are restructuring hard: consolidating divisions, and — in the words of one chief executive — exiting the staff whose skills can’t be reskilled “on a compressed timeline.”
Their most promising path is the one the frontier-lab post pointed at from the other side. The model providers need an enterprise distribution and integration channel; these firms are that channel — the ones who actually deploy agents into a regulated business and stand behind the result. Becoming the trusted, accountable deployment layer for agentic AI is a real and large opportunity. The catch is that their scale has always been measured in people, and if agents do the delivery, that scale flips from an asset into a liability — a lot of headcount to feed in a business whose unit of value is deflating.
The labour-arbitrage tier: most exposed
The traditional IT-services firms — the ones built on offshore headcount doing application development, testing, maintenance and support — face the sharpest version of this. Their model is the thing agents most directly automate, and the mechanism is brutally simple: the classic contract is priced on people and time, so if a smaller team delivers the same output, the link between headcount and revenue weakens — and so does pricing power. Analysts point to three pressures landing at once: fewer engineers needed per project, thinner wage premiums for routine skills, and shrinking project scopes as clients automate work in-house.
The market has already voted. These firms have been re-rated not because a single quarter was weak, but out of doubt that they can pivot from headcount-led growth to AI-led value fast enough. They have real and growing AI revenue — but it sits on top of an enormous legacy base that is deflating underneath it, and the race is whether the new outgrows the erosion of the old.
The pattern under all three: the hour is the seat
Everything above is one force refracted through three business models. The billable hour is to consulting what the per-seat licence was to software — a proxy for value that agents sever. So the same pricing migration is underway: buyers now overwhelmingly prefer outcome-based engagements to time-and-materials, and firms are shifting real portions of their fees onto outcomes — which forces them to rewrite partner compensation, because outcome revenue is lumpy and volatile in a way that hourly billing never was.
The pyramid becomes a diamond
The deepest change is structural. The consulting pyramid — a few partners on top, a wide base of juniors doing the automatable work — was never just a staffing chart. It was an apprenticeship engine: juniors did the grunt work, and in doing it, they learned to become the partners. When agents take the base, the economics say don’t hire the juniors — you can’t justify staffing ten when three people plus agents deliver faster. The pyramid compresses into something more like a diamond: a thin layer of juniors, a thick middle of seniors amplified by agents, the partners on top.
That’s efficient, and it’s also a time bomb. If you stop hiring and training the base, where does the next generation of senior judgment come from? Every tier is quietly betting it can solve that later. It’s the most important unsolved problem in the business, and almost nobody has an answer.
They become the thing they sell
The most interesting move is that, to survive, these firms are turning into the very thing this series has been describing. Deploying agent swarms internally, productizing their delivery into platforms, and pricing the outcome rather than the hour — that is service-as-software. The advisors are becoming vendors of the exact model the startups are using to attack them. The difference, and their edge, is what they already own that a startup has to earn: the client relationships, the domain trust, and the standing to be accountable for a result inside a regulated enterprise.
The verdict, and the survival path
It’s the same map as everywhere else in this series, at maximum intensity. Durable: judgment, trust, relationships, proprietary IP, and ownership of the outcome. Squeezed: commodity delivery priced by the hour. A firm that owns trust and IP and stands behind outcomes is on the safe side of the line; a body-shop reselling billable hours over commoditizing delivery is the wrapper — exposed exactly like any thin layer over a commoditizing input.
So the survival path is clear even if the execution isn’t: move up into judgment, IP and outcomes; become the governed deployment layer enterprises trust to install agents and answer for what they do; shift the price from bodies to results; and — the hard one — solve the broken apprenticeship pipeline before it becomes a talent cliff. And if you’re on the buying side, the instruction inverts: buy outcomes and accountability, not bodies and hours.
One caution against the easy narrative: this is not an extinction. These are deeply capitalised firms with the relationships, the balance sheets and the reinvention muscle that carried them through offshoring, cloud and digital. What’s happening isn’t the end of consulting; it’s a business-model transition — and, as with every other layer in this series, the winners and the laggards are about to diverge much further than the comfortable middle ever expected.
A companion to The Agentic Stack — a five-part series
Leave a Reply