From slide factory to outcome factory: why consulting economics are changing
AI breaks the day-rate convention. The question is no longer whether consultants use AI but whether firms can stop selling effort and start underwriting results.
For most of consulting’s modern era, the unit of value has been the day rate. A firm sold smart people for a slice of time, multiplied that slice by a project plan, and the bill arrived as a function of effort. Outcomes mattered to the firm’s reputation, but the meter ran on input. Almost every other piece of the consulting business model — utilization targets, leverage ratios, the partnership pyramid itself — flowed downstream from that one pricing convention.
AI breaks the convention. Not because consultants will stop being useful, but because the input the meter measures has become much, much cheaper. If the same slide that took a manager and two analysts a week to produce can now be drafted in an afternoon, clients will not keep paying a week’s worth of fees for it. Fee compression on time-and-materials work is no longer hypothetical, and most large consulting firms have responded with some version of the same pivot: stop selling effort, start underwriting results.
This is the second in a series. The first article argued that AI is changing what consulting was really selling: the analyst layer goes first, but the deeper shift is a barbell that compresses generic analysis and concentrates premium value on judgment, risk-bearing, and execution. This piece is about the commercial consequence of that shift: what happens to consulting firms when the labour they sell stops being scarce.
Why the hourly model survived as long as it did
Consulting kept time-and-materials pricing for so long because it solved a hard measurement problem. Most strategy work has long, fuzzy causal chains between recommendation and outcome: the recommendation is implemented partially, the market shifts, the CEO is replaced, attribution gets messy. Charging by the hour sidesteps the question of who-caused-what and prices the resource directly. Both sides understood the bargain: the client paid for the firm’s time and bore the outcome risk; the firm bore the delivery risk on the work itself.
That bargain works only when delivery is expensive. The day rate prices scarce, hard-to-replicate professional time. When AI compresses the time required to produce most consulting artefacts by five to ten times, the day rate either has to come down, which destroys margins, or the unit of pricing has to change. Most firms have figured out that the second option is the only survivable one.
Outcome pricing — what is actually happening
Bob Sternfels, McKinsey’s CEO, told HBR in its January–February 2026 issue that “about a third of our revenues total are underwriting outcomes” — deals where part or all of the fee is contingent on whether the recommendation actually delivers the promised result. He added that “my hope is that that crosses a majority of the revenues by the time I’m done being the global managing partner.” That is not a description of where the firm is. It is a commitment to where it is going. For a partnership whose previous public position was that strategy advice was too noisy to price on outcomes, it is a remarkable disclosure.
It is not just McKinsey. BCG, in its April 23, 2026 results announcement, reported $14.4 billion in revenue, with AI- and tech-related services now over 40% of total revenue and AI services specifically growing at 25% year-on-year. A meaningful slice of that is structured as managed services or productised offerings rather than time-and-materials engagements. Bain extended its worldwide Palantir partnership on March 26, 2026, specifically, in Bain’s own language, to harness “Palantir’s software and Forward-Deployed Engineers to achieve the end-to-end delivery of AI use cases from strategic plans through to operationalisation.” Outside the strategy firms, Harvey is selling legal AI services on two-tier pricing: fixed fees for the commodity transactional work, premium fees for genuinely strategic advice. Sierra, Bret Taylor’s company, prices its agent platform on outcomes: the customer is charged when the agent successfully resolves the issue, not when it tries.
The pattern across these examples is the same. The firm or vendor takes on a piece of delivery risk that used to sit entirely with the client, and prices the engagement to reflect that risk transfer. In exchange, they capture a larger share of the upside when the work performs.
This is structurally different from “AI makes consultants more efficient.” It is consulting moving from selling labour to selling results.
Strictly speaking, it is not consulting “moving” to outcome-based pricing. It is the strategy houses converging on a model that turnaround and restructuring firms have run for forty years. AlixPartners’ founder Jay Alix introduced incentive-based success fees and the chief restructuring officer (CRO) role in the 1980s; AlixPartners managing directors routinely sit as interim CFO or CRO with P&L authority and signing rights, not as advisors writing recommendations someone else implements. Sternfels, in the same HBR conversation, was wry about what the old model felt like from inside: “if it worked, that was because they were smart, and if it didn’t work, it’s because I didn’t implement.” Outcome underwriting closes that gap. What is genuinely new is the Big Three trying to learn the model. The structural question is whether a firm built on a pyramid of MBAs can absorb an operating model originally built around senior operators.
Service-as-software and the consultant as builder
The cousin of outcome pricing is what some VCs call “service-as-software”: packaging consulting methodology as a product that the client uses repeatedly, rather than a recommendation the client receives once and acts on themselves. The horizontal version is a cautionary tale. Tome cut 20% of its team in April 2024 to pivot toward enterprise sales, sunset Tome Slides in April 2025, and rebranded around “make deals, not decks” before the original team eventually spun out as Lightfield. Generic AI-replaces-decks businesses do not sustain.
The vertical versions, where deep methodology is packaged into a working tool for a specific industry or function, tell a different story. Harvey, the legal-AI company, was confirmed at an $8 billion valuation in December 2025 (a $160 million Andreessen Horowitz–led Series F) and was reported by CNBC in March 2026 to be at $11 billion, with more than 100,000 lawyers across 1,300 organisations using it, including 50 of the AmLaw 100. Hebbia, in institutional finance, claims to serve over 40% of the market; one bulge-bracket bank reportedly chose Hebbia over a frontier model lab on the basis that the lab’s team “had to explain what a CIM was” during the pitch. Vertical depth is what Tome did not have and what Harvey and Hebbia do: proprietary evals, domain context, codified methodology.
The cleanest articulations of why this matters are coming from outside the consulting firms themselves. George Sivulka of Hebbia argues that the goal is not “individual AI productivity” but institutional intelligence: the capacity to encode firm-level know-how in agents and workflows that survive any individual employee’s departure. Once firm-level methodology exists as software, the economics shift: the marginal cost of delivering the methodology to one more client approaches zero, and the firm captures recurring revenue rather than project revenue.
The implication for consulting firms is that the most valuable artefact a project produces may no longer be the deck. It may be the agent, the eval rubric, the workflow, or the operating tool the client keeps running after the engagement ends. That is a profound change in what a consulting deliverable even is. The consultant becomes part-strategist, part-product-builder, and the firm starts to look, incrementally, more like a software company that happens to start its sales process with a strategy conversation.
There is a paradox here the consulting industry has not yet faced publicly. If methodology can be productised, the consultants who build it have a stronger incentive to leave and run the product themselves than to stay inside a partnership that pays them for billable hours. François Candelon, who built BCG’s AI thought-leadership franchise, left in June 2024 to join Seven2, the French private equity firm, as a value-creation partner. The job is the same logic moved to a different platform: take AI methodology and apply it inside portfolio companies, where the upside is captured directly rather than shared out across billable hours. Other ex-McKinsey and ex-BCG operators have founded their own AI-services companies on the same calculation. No major firm has yet articulated a public answer to why the most product-capable people in the building should stay. Until they do, the firms will keep losing the people best positioned to lead the transition.
The forward-deployed engineer as the new consultant archetype
The most concrete commercial signal of the shift is hiring. OpenAI’s Frontier Alliance with McKinsey, BCG, Accenture, and Capgemini, announced on February 23, 2026, is openly described as a programme to move enterprises from AI pilots to production. Behind that is a quieter datapoint: The Information reported on February 5, 2026 that OpenAI itself was hiring hundreds of “Frontier Deployment Engineers,” its name for people whose job is to sit inside enterprise clients and make the AI work in real conditions. The pattern is not OpenAI-specific. A joint Indeed and Financial Times analysis found that forward-deployed-engineer postings rose more than 800% between January and September 2025. Sierra and Harvey rely heavily on similar profiles. Bain’s extended Palantir partnership institutionalises the pattern at the strategy-consulting end. And Accenture is hedging on both sides. It sits in the OpenAI Frontier Alliance while also running the Accenture Anthropic Business Group, announced December 9, 2025, with roughly 30,000 professionals being trained on Claude (Dario Amodei called it Anthropic’s “largest ever deployment”). The big systems-integration firms are not betting on a single model lab; they are positioning themselves as the layer that sits between any model and any enterprise.
The role, broadly, looks like this: someone with strategy or product instincts, technical fluency to actually build with AI, and the temperament to embed inside a client’s organisation for months rather than fly in and out for two-week sprints. The closest precedent is Palantir’s forward-deployed engineer model, which 8VC’s Joe Lonsdale and others have argued is now the template for AI-era enterprise services.
This is a meaningfully different job from “management consultant.” It is also a meaningfully different job from “software engineer.” The people who do it well tend to be hybrids: ex-MBB consultants who learned to build, ex-engineers who learned to navigate enterprise politics. They are not produced in volume by either traditional consulting recruiting or traditional engineering hiring, which is part of why salaries for the role are climbing fast.
Whether the established firms successfully cultivate this profile internally, or end up losing the talent to startups and to OpenAI itself, is one of the most consequential open questions in the industry.
Three stress tests for firms
To separate genuine model change from window dressing, three stress tests are worth applying to any consulting firm’s “we are reinventing for AI” narrative.
The first is whether the firm bears actual delivery risk. Outcome pricing is meaningless if the contingent portion of the fee is small enough not to matter, or if the success criteria are written so loosely that any plausible outcome triggers payment. The real version means a measurable share of revenue is genuinely at risk on outcomes the client controls. Sternfels’ “roughly a third” figure, if accurate at the contractual level rather than the marketing level, would be substantial.
The second is whether the firm has assets that compound. A firm whose AI strategy is “all our consultants now use Claude and ChatGPT” has acquired no compounding asset; the same tools are available to its clients. A firm whose AI strategy is the slow accumulation of proprietary evals, agent libraries, methodology codifications, and domain-specific tools has built something. McKinsey’s Lilli, BCG’s GENE and Deckster, KPMG’s Workbench, Deloitte’s Zora are all plays for the second option. Whether they actually compound, or are essentially internal copies of capabilities clients can buy directly, is a question for the next twenty-four months.
The third is whether the firm has reduced its dependence on leverage ratios. A firm whose economics still require a 6:1 analyst-to-partner pyramid to generate partner profit has not yet adjusted to a world where AI compresses the analyst layer. The harder transition, to a smaller, more senior, more technical workforce, is operationally and culturally painful.
AlixPartners describes its own delivery model as “senior-led, cross-functional”; the firm’s recruiting language is blunter: “we have more mentors than mentees.” Industry observers have described the resulting shape as a “diamond”: a thin junior intake, a thick experienced middle of managing directors who are often former operators (CFOs, COOs, turnaround specialists), and a smaller partner apex. Whatever you call the shape, it is the structure an outcome-underwriting model actually requires. AlixPartners’ own 2026 Disruption Index, released January 14, 2026, surveyed 3,200 executives across 11 countries. It found 51% of self-identified growth leaders had widely implemented agentic AI, versus 14% of laggards. Separately, 57% of CEOs expect significant business model change in the next year. As Rob Hornby, the firm’s UK-based co-CEO, put it: “Expectations around AI integration are redefining the C-suite agenda.” The full disclosure is that I work inside this model; the observation worth disclosing is that outcome delivery requires people who have signed P&L statements, not just people who have built models of P&L statements. The question for McKinsey, BCG, and Bain is not whether they can sign more outcome-based contracts. It is whether a firm whose economic engine is leveraging armies of MBAs can be re-shaped without breaking partner economics.
The honest counter-case
It is worth flagging the case against this whole picture, because it is more credible than the AI-bullish version of the story usually admits.
Outcome pricing has been promised in consulting for thirty years. Most attempts have failed because attribution is genuinely hard and because both sides eventually preferred the cleaner accounting of time-and-materials. The current wave may go the same way. Service-as-software has run aground repeatedly when methodology turned out to be less codifiable than expected, or when the productised version commoditised faster than the firm could capture the upside. The forward-deployed engineer model is expensive and culturally hard for firms whose hiring and promotion systems were built around generalists.
And the macro picture remains stubbornly mixed. The MIT NANDA finding that 95% of enterprise AI initiatives produce no measurable P&L impact is not consistent with a world where consultants have figured out how to underwrite outcomes at scale. NBER work showing roughly 90% of CEOs report no measurable productivity or employment effect from AI over three years cuts in the same direction. McKinsey’s Joe Ngai, the firm’s Greater China chair, told Caixin Global on April 14, 2026 that 90% of clients are experimenting with AI, 40% are seeing some financial return, and the AI contribution to EBIT remains under 5% even at firms in the success category.
So the honest version of the prediction is not that consulting is being transformed in real time. It is that the most credible firms are actively trying to transform it, the commercial logic is pulling in the right direction, and the question is whether the transformation arrives faster than the underlying hourly business shrinks. Two firms can pursue identical strategies and one can succeed and the other not, depending on whether the pivot lands before the billable backlog erodes.
The slide factory is not dead. It is being asked, for the first time in fifty years, to prove it deserves what it charges.
Prediction tracker
Claim: By the end of 2027, at least one Big Three firm reports, at the contractual rather than the marketing level, that more than 40% of its revenue is structured on outcome-contingent terms.
Confidence: Low to medium. Sternfels has stated the intent; the question is execution, partner-comp math, and whether the contracts contain real risk transfer rather than loose triggers.
Falsifier: If, by end-2028, the Big Three’s disclosed outcome-contingent share is still hovering around the current “roughly a third” or has been quietly redefined to include any fee with a bonus clause, the conversion-layer-to-outcomes thesis is wrong about pace, even if right about direction.