There is a $1.5+ trillion prize for rebuilding how the enterprise actually works. Many of the firms that dominated the last era may just not be there to collect it.
Start with the size of the prize, because almost everyone is underestimating it.
The opportunity in front of the services industry is not a software refresh and it is not an “AI strategy” slide in a board deck. It is a once-in-a-generation rebuild of the enterprise itself, running on two tracks at once. The first track is the slow, expensive refactoring of the legacy systems of record, the ERPs and claims platforms and core banking systems that were designed for humans to type into. The second track is brand new, and it is where the real money lives: building systems of work, the semi-autonomous operations that do not just store what a company knows but actually execute what it does.
HFS Research, which coined the term Services-as-Software, puts the number at $1.5 trillion by 2035, a market that grows by absorbing revenue from traditional IT services as that pool shrinks and from software as it slowly evolves. Sequoia’s Julien Bek frames the same shift from the demand side with a ratio worth memorizing: for every dollar a company spends on software, it spends roughly six on the services and labor that make the software actually do something. The last generation of enterprise vendors fought over the software dollar. The next generation is going after the other six. That is not a bigger slice of the same pie. It is a different and much larger pie, and it is made mostly of services payroll intertwined deeply with AI Software.
This is the largest reallocation of enterprise spending in forty years. The question is not whether it happens. The question is which firms are standing on the right side of it when it does.
The decoupling nobody on a bench wants to discuss
Frank D’Souza, who ran Cognizant and now invests in digital services firms, has put the underlying mechanic about as plainly as it can be put: “the value of code is infinite, but the value of coding is zero.” The output, working and trustworthy software, has never been more valuable. The act of producing it has never been cheaper. Those two values, fused together for four decades, are now coming apart.
AI is the new India, and the comparison is exact. Labor arbitrage was the last great operating-model disruption, the move that let firms staff armies of engineers in low-cost geographies and bill them by the hour. AI is the same kind of disruption pointed at the same cost structure, except this time the cheap labor is not in another country. It is in a model. Everest Group’s Peter Bendor-Samuel, who watched the offshore model take hold from its earliest days, has been blunt that AI strikes directly at the economic engine of the traditional firm: the talent pyramid that monetizes implementation, maintenance, and process management by keeping a wide base of junior people busy under a thin layer of seniors.
Here is the part the industry keeps trying not to say out loud. The pyramid was never just a cost structure. It was the firm’s entire theory of how work gets done and how quality gets guaranteed. Juniors do the work, seniors review it, knowledge transfers upward over years, and the client pays for the whole apparatus. The pyramid was a magnificent machine for converting junior salaries into senior margin. AI just installed a cheaper machine, and nobody bothered to tell the org chart. It does not trim the bottom of the pyramid. It removes the floor the building was standing on, which is precisely why treating AI as an “efficiency overlay” on the existing model is a category error. The shape that replaces the pyramid is a diamond: a thin base of curated apprentices, a broad middle of orchestrators and architects and accountability stewards directing fleets of agents, and far fewer hands than before.
Why the J-curve is lethal, and only to some
Every transformation has a J-curve, the dip before the climb. For a services firm the dip is brutal and specific. AI cuts the hours you can bill before your new outcome-based and IP-based revenue has grown large enough to replace them. Revenue sags, then, if you survive, it recovers and compounds. It is the single fear that surfaces most in any honest conversation among services leaders, and for good reason.
The trap is not that incumbents cannot see the J-curve. They can read the map perfectly well. They just cannot afford the toll. Everything about how a large, successful services firm is built punishes the dip with maximum force. Quarterly expectations from public markets. Utilization targets that treat an idle consultant as a small tragedy. Partner compensation pools tied to headcount and contract volume. Decades of headcount-centric muscle memory, where sales leaders pitch capacity instead of certainty and delivery leads instinctively solve a hard problem by throwing more people at it. A firm optimized to win the labor era is, by construction, optimized to refuse the very dip the AI era demands.
So the leap is not intellectually hard. It is structurally hard, and it is hardest for exactly the firms that were best at the old game. The more successful you were at scaling billable headcount, the heavier your anchor when it is finally time to cut loose from it.
This is where the J becomes the K.
The K has already started to form
Picture two firms standing at the same point in 2025. One commits to the dip, rebuilds its commercial model around outcomes, and starts compounding proprietary data and agents. The other defends the pyramid, protects this year’s margin, and quietly optimizes itself into a slow decline. From that shared starting point the trajectories split, one arm bending up and one bending down. That is the K-curve, and the cruel part is that both firms look perfectly healthy in the brochure right up until the moment the divergence shows up in the cash flow.
The numbers are starting to arrive. Tata Consultancy Services, a bellwether of the entire offshore model, posted its first revenue decline since going public, which HFS read not as a company stumble but as a leading market indicator. The broader outsourcing market has been decelerating, and the slowdown is concentrated, as you would expect, among the less dynamic firms that have the most to protect and the least appetite to cannibalize it. The valuation gap tells the same story in starker terms: in early 2026 the market valued OpenAI at roughly $840 billion and Anthropic at around $380 billion, while the publicly listed technology services firms watched their valuations slide over the prior twelve months. Capital is pricing in a world where leverage comes from code, data, models, and agents, and not from talent-heavy delivery.
The lesson compresses into one sentence. If you do not act on the J, you will spend the back half of this decade on the bottom arm of the K.
The Jetson Index
George Jetson’s entire job was pushing a single button, and he still came home grumbling that work was exhausting. That cartoon got the future of work exactly right. The labor does not vanish. It moves up the stack, from doing to directing, and the firms that win are the ones built around the person pushing the button rather than the thousands who used to do the job by hand.
So here is a way to tell, without waiting for a press release, which arm of the K a services firm is actually on. Call it the Jetson Index, five questions whose honest answers are very hard to fake.
The first is decoupling. Is revenue still handcuffed to headcount, or has revenue per employee broken free of the hiring plan? Growth that still requires linear hiring is just a pyramid wearing a lanyard.
The second is IP density. How much of delivery runs on proprietary, reusable assets, the code and data and models and agents that constitute a real system of work, versus labor rebuilt from scratch on every engagement? The test is simple: does each project lower the cost of the next one, or reset the meter to zero?
The third is pricing. What share of revenue comes from outcomes and guaranteed results rather than time and materials? Billing by the hour while the hours evaporate is not a business model, it is a countdown.
The fourth is shape. Pyramid or diamond. Count the FDEs, orchestrators and architects steering agents, count the juniors billing hours, and notice which number leadership is actually trying to grow.
The fifth is the dashboard test. What single metric does the executive team genuinely manage by? If the answer is utilization, the firm is optimizing beautifully for an era that is ending.
Score a firm honestly across those five and you will know more than most quarterly reports will tell you. The catch is the word honestly. Most incumbents will grade themselves generously, the market will grade them accurately, and the deal terms will settle the argument. Or perhaps given the importance, have an external agency do the rating and continued validation.
The proof of concept is already public, and it is called Palantir
If you want to know whether the consulting-led, outcome-driven, data-grounded model actually works, you do not need a forecast. You need Palantir’s earnings.
Palantir reported 85% year-over-year revenue growth in the first quarter of 2026, its fastest since going public, and returned something on the order of 640% to shareholders over five years. It did this with a model the rest of enterprise software spent a decade refusing to copy because it looked too expensive and too weird: the forward-deployed engineer (FDE), embedded inside the client, wiring the company’s messy real-world data into an ontology and standing up agents that take action against it. Palantir’s own framing is that its platform is an operational system for deploying agents with governance, provenance, and auditability, not a thin wrapper around someone else’s model. This seemingly strange company at the enterprise software party now owns the house. And the CEO, in all his quirkiness, clearly has style in doing so as he frequently sports a Patek Aquanaut ref 5968A-001 with an orange band. Well worth a K+ rating.
The structural magic is what separates this from classic consulting. A traditional firm builds something once for one client and bills the hours. Palantir builds it once, watches how it breaks, and folds the lesson back into the platform, so every engagement is effectively R&D that lowers the cost of the next one. The field work generates the product. The advantage compounds instead of resetting at the start of each project. That is the upper arm of the K rendered as a public company, and its CTO Shyam Sankar summarized the era in seven words: “Tokens are the new coal. Palantir is the train.”
This is the model that owns the systems of action or work: grounded in proprietary data, sold as an outcome, accountable for the result. It is also the model the most powerful new entrants in the market have decided to copy wholesale.
The frontier labs have stopped staying in their lane
For two years the comfortable assumption among services firms was that the model makers would sell the engine and leave the coachwork, the integration and deployment and accountability, to the partners. That assumption is dead. The engine makers looked at the coachbuilding business, did the math, and decided to build the entire car.
The frontier labs are moving up the stack into the application and services layer, and they are doing it with balance sheets the incumbents cannot match. OpenAI launched the OpenAI Deployment Company with more than $4 billion of initial investment, structured as a roughly $10 billion vehicle anchored by TPG with nineteen investment firms behind it, and its strategy is explicit: turn the portfolios of large private equity houses into a captive distribution channel. To get delivery muscle on day one it agreed to acquire Tomoro, an applied AI consulting firm, walking roughly 150 forward-deployed engineers in through the door. Within days, Anthropic announced its own enterprise services joint venture valued at $1.5 billion with Blackstone, Hellman & Friedman, and Goldman Sachs as founding partners. Google is playing both sides, offering direct equity investment to startups through its AI Futures Fund while committing $750 million to push partners deeper into agentic delivery. Reuters has reported that both leading labs are actively hunting for services firms to buy.
Read that pattern carefully. The companies with the best models on Earth looked at the services market, concluded the deployment layer was where the value and the lock-in lived, and chose to build and buy their way into it rather than wait for the incumbents to modernize. They are copying Palantir’s forward-deployed playbook on purpose. The land grab is not a future risk. Per Menlo Ventures, Anthropic alone moved from roughly a quarter of the enterprise AI market at the start of 2025 to around 40% of it, and the major integrators have collectively committed more than a million practitioners to deliver on these platforms. The window in which a labor-based firm could treat AI as someone else’s product has quietly closed.
The money has already picked a side, which is why the deal market is about to get loud
The most telling signal is not what the labs are saying. It is what private equity is doing.
The same investors who, in the last cycle, would have bought into a Big Four transformation practice or rolled up a regional systems integrator are now financing the replacement. The people who used to buy the pyramid are now writing checks to the wrecking crew. When Blackstone anchors an Anthropic services venture and TPG underwrites OpenAI’s, the verdict is in: sophisticated capital has decided the fastest route onto the upper arm of the K is to build or buy something AI-native, not to wait for a pyramid to reinvent itself. This cuts two ways at once, and most commentary only notices one of them.
The first is private equity as a new entrant, capitalizing AI-native delivery to wring margin out of its own portfolio companies. That is net-new competition arriving from a direction nobody was watching. The second, and the more interesting one for anyone holding the right asset, is private equity as an acquirer. The same logic that makes OpenAI pay to absorb 150 deployment engineers makes any genuinely AI-native services firm with real domain depth a scarce and strategic prize. When the labs, the hyperscalers, and the buyout funds are all competing to acquire deployment capability they cannot build fast enough, the firms that already have it stop being vendors and start being targets.
Expect the deal market to go into a frenzy for “true” AI-native services and solution companies. The criteria that will command a premium are easy to predict, because they are the precise opposite of what the old model optimized for: proprietary data and IP rather than benchcount, outcome accountability rather than utilization, and ownership of the end-customer relationship rather than subcontracted invisibility. In the labor era, headcount was the moat. In the AI era, it is the discount. Nobody is paying a premium for a warehouse full of billable hours.
Do not mistake the gap for a reprieve
There is a catch, and it is the reason this is a strategy paper and not an obituary. The technology is ready before the buyers and the regulators are.
Enterprises move slowly, especially in banking, insurance, and healthcare, where a wrong answer is a regulatory event and not just a bad ticket. The bottleneck has simply moved. The cost of producing code has collapsed, while the cost of certainty, the work of making output secure, compliant, explainable, and integrated, has gone up. Somebody still has to stand behind the result when an agent gets it wrong. That accountability gap, plus procurement inertia and a still-forming regulatory picture, buys the industry a window of perhaps two to three years before the new model becomes the default rather than the exception.
That window is real. It is also a trap if you read it as time to relax. Two to three years is not a moat. It is a hall pass. The pace will slow. The direction will not change. The writing here is not on the wall in pencil. It is carved in hieroglyphics, deep and permanent, and it is not going to weather off because a few buyers are nervous and a few rules are still unwritten. The window is for getting onto the right arm of the curve while entry is still cheap, not for admiring the view from the wrong one.
The only question that matters now
For the leaders of services firms, the question has changed. It is no longer “how do I produce more code,” but how do I turn abundant code into trusted outcomes faster than anyone else, without losing integrity or security. That is a different business with a different shape, different metrics, and a different kind of person at its center.
For the investors and buyers watching from the outside, the discipline is harder, because the K-curve pays for conviction held before the divergence is obvious. By the time the two arms have visibly separated in the public numbers, the entry price on the winners will already reflect it. The firms worth backing are the ones that score high on every line of the index before it is fashionable: that never had a pyramid to dismantle, that treat domain depth and proprietary data as the product, and that put the human in the chair of the accountable steward rather than the billable unit.
The labor era rewarded the firms that could add the most people. The era now beginning will reward the firms that need the fewest. Everyone in the market can see the same future. Only some of them are built to survive the trip there.
Jim Francis, Technology Futurist & Optimist, CEO / Founder ConceptVines Contact: linkedin.com/in/jimgfrancis, www.conceptvines.com