The Consulting Pyramid Is Reshaping Around AI
AI is automating the analyst layer that consulting has run on for fifty years. Here is what the structural response looks like at the firms building it first.
The consulting pyramid runs on a simple premise. Partners hold the most valuable asset in the firm: relationships and judgment. Analysts hold the cheapest one: time. The model works because partners can sell work at a high margin and then deploy analysts to execute the research, modelling, and synthesis that make the engagement billable. One partner can support ten analysts. Revenue scales without proportionally scaling senior headcount.
AI is dismantling the economics of that model, faster than most firms expected.
What AI automates in a consulting engagement
Junior analyst work in consulting is largely knowledge work performed at high volume: data collection, market benchmarking, financial modelling, regulatory summarisation, slide assembly. These tasks require discipline and competence, but not the senior judgment that clients pay the highest rates for. They are precisely the category of work that large language models, paired with retrieval and tool use, now perform at speed.
McKinsey's internal AI assistant, Lilli, is used by over 72% of McKinsey's global workforce and has reduced research and synthesis time by around 30%. BCG's Deckster builds presentation decks in minutes. Bain's Sage copilot is trained on internal IP and accessible across client teams. These are not edge-case tools or proof-of-concept pilots. They are live deployments at scale, compressing the hours that junior consultants previously billed directly to clients.
The staffing impact is now measurable in public data. McKinsey cut roughly 3,000 to 4,000 positions in 2025 and 2026, with reductions concentrated in junior research and back-office functions where AI has had the largest productivity effect. Across the Big Four, graduate intake has dropped sharply: KPMG reduced its graduate class by 29%, Deloitte by 18%, EY by 11%, and PwC by 6%. Overall, graduate job postings at the major accounting and consulting firms are down 44% in 2025 compared to the prior year.
The direction is consistent. The analyst layer that made the pyramid work is being compressed by AI productivity, and the firms that can move fastest are doing so.
The structural response at the top of the market
The shift is most visible in how the largest firms are reorganising themselves, not just their headcount.
In January 2026, Deloitte announced it would scrap traditional job titles for approximately 181,500 U.S. employees, effective June 1, 2026. The analyst-consultant-manager hierarchy, which mapped directly to the layers of the consulting pyramid, is being replaced by role-specific functional titles that describe what people actually do. An employee formerly called a "consultant" might become a "Software Engineer III" or a "Senior Consultant, Functional Transformation." The internal promotion ladder, previously expressed in opaque seniority designations, is becoming explicitly skill-and-function-based.
The September 2025 HBR article "AI Is Changing the Structure of Consulting Firms" described the likely structural successor to the pyramid as the "obelisk": fewer layers, smaller teams, with senior engagement architects leading work that AI agents handle the analytical execution for. The obelisk preserves senior judgment and client relationships at the top while removing much of the junior scaffolding underneath.
A related model is the "diamond": wider in the middle than at the base, with a substantial tier of experienced specialists and AI orchestrators replacing the broad base of entry-level analysts. The logic is similar to the obelisk but places more emphasis on building out the middle tier, where people who understand both domain and AI tooling become the primary delivery vehicle. Both models converge on the same structural conclusion: the pyramid's leverage came from stacking junior headcount; the new leverage comes from AI infrastructure underneath smaller, more experienced teams.
Platform-and-people replaces people-heavy delivery
The deeper structural shift is about the nature of what gets delivered, as much as who delivers it.
When AI handles the analytical labour, the value delivered in an engagement can no longer be justified on hours spent. Clients increasingly expect to see systems they can access and measure, not reports they need to act on manually. The evidence is in how the largest firms have repositioned their offerings: as platforms with advisory wrapped around them, rather than advisory labour supported by analysts underneath.
Deloitte's CortexAI is a cloud-enabled platform with plug-and-play datasets, analytics dashboards, and AI capabilities, positioned as continuous access to analytical infrastructure rather than a project deliverable. McKinsey's Periscope combines analytics tools with expert support and training. Its build-operate-transfer framing makes the delivery model explicit: the firm builds a system, operates it alongside the client, and eventually transfers ownership. Accenture reported $5.9 billion in generative AI bookings for fiscal year 2025, with AI revenue reaching $2.7 billion, roughly three times the prior year. Its AI Refinery platform and industry AI playbooks follow the same structural pattern at larger scale.
These are not experiments or future-state ambitions. They represent the delivery model the top of the market has settled on: platforms built on AI infrastructure, delivered by leaner teams, sustained through recurring access rather than project closure. The revenue logic has shifted from billing hours to building assets.
What this means for mid-market and specialist firms
The transformation at McKinsey, Deloitte, and Accenture sets the competitive context for every firm below the top of the market.
When the largest firms reduce their analyst base and invest in platforms, they can service more clients with smaller teams. Mid-market firms that still run a deep junior pyramid are now competing on cost with firms that have AI infrastructure doing the analytical work. The proximity advantage that smaller boutiques once held, more senior attention per engagement, erodes if clients can get comparable analytical output from a platform-backed large firm for less.
The structural response for mid-market and specialist firms mirrors what the large firms are executing: build platforms that retain and leverage IP across engagements, shift team composition toward specialists and AI orchestrators, and move pricing toward fixed-fee packages and recurring access rather than time-and-materials. The scale of investment differs, but the structural logic does not.
The operating unit that makes this work in practice is what Gartner calls the fusion team: a persistent, cross-functional group that owns a deliverable across its lifecycle, generates telemetry from client usage, and iterates based on what the system surfaces. A fusion team typically includes domain expertise, UX, data and analytics, an application builder, and someone accountable for adoption. This structure is incompatible with a pure pyramid because it requires sustained ownership rather than project-by-project staffing rotations.
The leverage model has shifted
The consulting pyramid's leverage logic was always about one senior person unlocking the capacity of many junior ones. That logic held as long as junior work required a team of humans to produce. AI has collapsed the marginal cost of that layer.
What replaces the pyramid is a delivery model built on platforms, specialist orchestration, and recurring access. The leverage still exists. It has moved from people stacked below partners to software and AI running beneath smaller, more experienced teams. Firms that recognise this early have an execution window before mid-market clients begin to expect platform-shaped outputs as the default.
The shift in what gets delivered, from documents to interactive systems, and the shift in how delivery is staffed, from pyramids to fusion teams, are the same transition viewed from two different angles. We explored the deliverable side of that shift in Decks Are Dying: Why Service Deliverables Are Becoming Software. The structural and operating model implications are where it gets operationally real for firm leaders.
For a structured diagnostic of where your firm sits on this spectrum today, our Service Productisation Assessment playbook maps the transition from project-based delivery to productised operating models. For the build-and-deploy layer, Gleo is the platform Beach uses to assemble and run interactive delivery artefacts across its own client programs.