In November 2025, KPMG UK put its first 134-person cohort, early-career staff through to Advisory partners, into an AI upskilling program delivered by Multiverse, a specialist provider. One of the four largest professional services firms on earth bought AI training from somebody else. That is the most useful fact in this comparison, and it sits in a vendor press release rather than a buying guide.
A disclosure first. SFAI Labs runs facilitated AI enablement sessions in which teams work with assistants on their own documents, so we compete with part of what follows. We also tell prospects to hire a Big Four firm when their problem is a regulatory one, and this says when that is.
Two different things are called Big Four AI training
Search the phrase and you get a workforce story. Deloitte’s PairD platform reaching 75,000 staff across Europe and the Middle East. PwC’s three-year, one-billion-dollar US commitment from April 2023. EY’s 1.4 billion dollar investment announced alongside the EY.ai platform in September 2023. KPMG’s alliance with Anthropic, which Anthropic’s own newsroom describes as giving every one of KPMG’s 276,000-plus employees access to Claude across 138 countries and territories.
All true, all irrelevant to a purchase order. Those programs exist because the firms’ revenue depends on billable people who can use these tools. They are evidence of seriousness, not of fit.
The second meaning is the one you can buy: AI literacy and enablement sold as a service line. That business is growing for reasons that have little to do with productivity. The EU AI Act’s Article 4 obliges providers and deployers of AI systems to ensure a sufficient level of AI literacy among staff and others operating systems on their behalf. The duty has applied since February 2, 2025, and the European Commission’s position is that supervision and enforcement begin on August 2, 2026. No certification is required, though the Commission encourages keeping a record of what you did. Compliance-driven demand is the best kind, and the Big Four organized around it first.
Only one of them publishes a price, and only at the cheap end. A PwC Academy e-learning course on AI literacy and governance is listed at £100 per learner with 12-month access on ACCA’s CPD finder, and PwC Netherlands priced a blended board-level session at €395 a seat in November 2025. Facilitated engagement pricing is published nowhere.
What each firm actually sells to clients
Read the firms’ own pages rather than the trade coverage and the offers separate.
| Firm | Client-facing offer | What its own pages say |
|---|---|---|
| Deloitte | Academy for AI, the external version of its internal AI Academy | 30-plus courses across four areas. The role-based paths named on the page are Agentic AI Engineer, GenAI Engineer and Data Scientist, plus trusted-AI governance. |
| PwC | PwC Academy, by country | AI literacy e-learning curricula pitched explicitly at the EU AI Act duty, plus board-level sessions and tailored programs. |
| KPMG | AI Learning Academy, with a free Skills Development Centre as an on-ramp | An enterprise AI learning and enablement program for leaders and the workforce. Country-specific, with Canada’s the most developed publicly. |
| EY | EY.ai-branded learning and badging | Built around its own platform and a badge system covering roughly 400,000 employees. |
Deloitte markets the Academy for AI on the basis that it trains its own workforce with the same material, and publishes the proof: every US Deloitte professional took at least one AI training in the past year, 300,000 trainings completed. That is a genuine credential and the strongest argument any of the four makes.
Then look at what the role paths are called. Agentic AI Engineer. GenAI Engineer. Data Scientist. Those are curricula for people who build systems, not for a controller who needs a usable first draft of the monthly variance commentary without pasting board material into a public chat window. The same mismatch runs through the hyperscaler catalogs, which is why the free Microsoft Learn paths get resold at a per-seat premium with a facilitator attached.
The only published window into Big Four pricing
Commercial rates are confidential. Federal ones are not. Deloitte Consulting LLP’s price list for its General Services Administration Multiple Award Schedule contract GS-02F-0130W is posted on deloitte.com, and Talent Development is one of the special item numbers in its scope. The final contract year in that document, March 4, 2024 to March 3, 2025, carries these hourly ceiling rates for its HR technology labor categories:
| Category | Published ceiling rate |
|---|---|
| Principal / Managing Director III | $368.65 |
| Senior Manager III | $322.41 |
| Manager I | $281.95 |
| Senior Consultant III | $266.45 |
| Senior Consultant I | $188.00 |
| Consultant I | $150.17 |
| Business Analyst I | $125.52 |
Read the caveats before quoting these to anyone. They are ceiling rates on a government schedule, discounted off commercial pricing, and that contract period closed in March 2025. Commercial day rates are higher and unpublished. The numbers are a shape, not a quote.
The shape is the point. The published grades span nearly a factor of three, and a firm that bills $368 an hour at the top and $125 at the bottom does not make its margin on the top row. It makes it on the mix.
The staffing pyramid is the product
Almost every Big Four engagement is sold by someone who will not deliver it. A partner builds the relationship, a senior manager writes the statement of work, and delivery lands on consultants and analysts who are capable, hard-working, and frequently two years out of college. The pyramid is not a scandal; it is how the firm trains its own people.
It is also the variable that decides whether your AI enablement works. A room of skeptical department heads tests the facilitator inside fifteen minutes. They ask something specific about their own workflow, and either the person at the front has done that work or has not.
So the question to put in writing is not “do you have AI expertise.” It is: name the people who will facilitate each session, state their grade, and say how many sessions of this kind each ran personally in the last twelve months. Then put the names in the contract. Our guide to the clauses that decide whether training lands covers how to make that stick.
A specialist cannot play this game, because there is no bench to hide behind. Whoever sold the work is in the room on Monday: a structural advantage at your size, a structural limitation at enterprise scale.
What a 200-person company is to a firm with a £456 million contract
In August 2026 the UK Cabinet Office awarded KPMG and EY civil-service training contracts worth up to £456 million in total, split roughly £319 million to KPMG and £137 million to EY and running to 2028. It is a ceiling rather than guaranteed spend, and the data provider Tussell called it the largest award to Big Four firms in its records since 2012.
Hold that next to a $40,000 enablement program for 180 people. Both are real revenue, but only one decides who gets the best facilitators in November. Staffing at a pyramid firm follows account priority, and the symptoms below a firm’s target size are familiar: a strong pitch team, a different delivery team, a change request priced as a new engagement.
That is not a reason to dismiss the Big Four, but a reason to ask where you sit in their portfolio. If your spend with the firm already has another zero on it from audit or tax work, the calculus changes in your favor.
Where a specialist wins, and the evidence for it
The strongest case against brand is not a complaint about consultants. It is BCG’s own 2025 research, a survey of more than 10,600 leaders, managers and frontline white-collar employees across 11 countries and regions:
- Regular AI use among frontline employees has stalled at 51 percent, while more than 75 percent of leaders and managers use generative AI several times a week.
- Only one-third of employees say they have been properly trained.
- Regular usage is sharply higher for employees who receive at least five hours of training and who have access to in-person training and coaching.
Every variable in that third finding is a format variable. Hours, live instruction, follow-up coaching. None is a function of the logo on the invoice. A large firm can deliver all three and a specialist can fail at all three. Buy against those variables instead of against a brand, the same discipline our case that licenses are not capability applies to the software line.
Where a specialist typically wins: specificity, because the session runs on your documents rather than a generalized catalog. Facilitator identity, because the seller delivers and you can verify the person before signing. Speed, with a scope in days and a first session in weeks. And price, since facilitated workshops sit in roughly the $2,000 to $15,000 range depending on headcount and session count, which is where large-firm engagements start.
Where a specialist loses: multi-country regulatory sign-off, an audit committee that wants a recognized name, and procurement with a vendor revenue floor.
The hybrid most companies should buy
The framing of this comparison is slightly wrong, and saying so saves money. Big Four AI advisory and facilitated enablement are not substitutes; they are two purchases bundled by whoever wrote the first proposal.
One is governance work: an AI policy, a risk register, a control framework, an Article 4 literacy record that survives a regulator’s question. Large firms are good at this across jurisdictions, and the deliverable carries weight partly because of the letterhead.
The other is capability work: getting 180 people to produce better output on Tuesday than on Monday. That is measured in hours of live practice on real work, and its quality depends on who stands in the room.
Buy them separately. The unhappiest Big Four training clients bought a policy engagement and expected a capability outcome, then found at the readout that the training module was recorded content appended to a governance project. Our buy-versus-build analysis breaks the same decision down asset by asset, because curriculum, facilitation, measurement and maintenance each have a different right answer.
A scoring sheet for the decision
Score each row honestly. More marks on the left means the premium buys something.
| Question | Points to Big Four | Points to a specialist |
|---|---|---|
| Do you operate AI systems in the EU or another regulated jurisdiction? | Yes | No |
| Does a board or audit committee need to see a recognized name? | Yes | No |
| Does procurement require a vendor above a revenue floor? | Yes | No |
| Is the deliverable a defensible document, or changed behavior? | Document | Behavior |
| Do you already spend seven figures with the firm on other work? | Yes | No |
| Can the vendor name your facilitator in the contract? | Rarely | Expect it |
| Do you need the first session inside six weeks? | No | Yes |
Whichever way it lands, verify the proof rather than the pitch. Every vendor here leads with a case study, and a case study is an advertisement rather than evidence. Ask for a reference call with a company your size, then ask that company who facilitated.
A thirty-day sequence
- Week one. Decide which of the two purchases you need, and write two scopes if the answer is both.
- Week two. Send the same five questions to one large firm and two specialists: who facilitates, at what grade, how many comparable sessions in the past year, what material the first session uses, and what you keep. Compare answers, not decks.
- Weeks three and four. Run a paid pilot with one department, then decide on evidence: scale the format that worked and keep the governance scope separate so you can renew the two independently.
If you want a second opinion on which purchase you need, we run a free AI-readiness call. No deck, and we say so when the answer is a governance engagement rather than a workshop.
Frequently asked questions
Do the Big Four sell AI training to outside companies, or only train their own staff?
Both. Deloitte sells the external version of its internal academy as Academy for AI, PwC sells AI literacy curricula through its country Academy entities, KPMG runs an AI Learning Academy, and EY builds client learning around EY.ai. Only PwC publishes per-seat prices, for e-learning.
How much does Big Four AI training cost?
No firm publishes commercial rates. The closest public reference is Deloitte Consulting’s General Services Administration schedule price list, where hourly ceilings in the final year of contract GS-02F-0130W ran from $125.52 for a business analyst to $368.65 for a principal. Those are discounted federal ceilings, and the period ended in March 2025. Expect a large firm’s facilitated program to start above the $2,000 to $15,000 range specialist workshops occupy.
Is a Big Four firm better at AI training than a specialist?
They are better at different things. Large firms are stronger on multi-jurisdiction regulatory work, board-level credibility, and bundling training into a governance scope. Specialists are stronger on facilitator quality at a given price, speed, and building the session around your own documents. BCG’s 2025 research ties regular use to training hours and in-person coaching, not vendor size.
Why did KPMG UK hire Multiverse to train its own people?
Multiverse announced in November 2025 that KPMG UK’s first AI upskilling cohort, 134 people from early-career staff to Advisory partners, was enrolled in a Level 4 AI for Business Value program leading to a formal qualification. KPMG has not published its reasoning. The straightforward reading is that delivering structured upskilling at scale is a specialist capability, worth remembering when a large firm offers to sell you one.
Does the EU AI Act require us to buy AI training?
Article 4 requires providers and deployers of AI systems to take measures ensuring a sufficient level of AI literacy among staff and others operating systems on their behalf. It has applied since February 2, 2025, with Commission supervision and enforcement starting August 2, 2026. It mandates no vendor and no certification. The Commission has encouraged keeping a record of your AI literacy efforts, so documentation matters more than the logo on the course.
What should we ask a large firm that a specialist would answer easily?
Name the facilitators in the contract. Specialists expect the question because the seller delivers the work. If a large firm commits a named individual with a history of comparable sessions, most of the concern here goes away.
Can we use the Big Four for governance and a specialist for the workshops?
That is the arrangement we see work most often. The governance scope produces a policy, a risk register and a literacy record; the capability scope produces hours of facilitated practice. Contract them separately so either can be replaced independently.
What if our board insists on a Big Four firm?
Give them the brand on the governance work and negotiate facilitation separately. If the whole program must sit with one firm, make the named-facilitator clause the price of the signature and scope a paid pilot session before committing the rest.
Key takeaways
- “Big Four AI training” means two things: how the firms train their own hundreds of thousands of employees, and the academy service line you can buy.
- Deloitte’s role paths are named for engineers and data scientists. If you need business-user fluency on everyday documents, check the catalog you are shown was built for that.
- No firm publishes commercial pricing. Deloitte’s federal schedule rates, $125.52 to $368.65 an hour in the final year of a contract that ended in March 2025, are the only public window, and the spread across grades is the business model.
- The decision is about who delivers the hours. Ask for named facilitators, their grade and their session count over the past year, then put the names in the contract.
- Format beats brand. BCG ties regular use to at least five hours of training and in-person coaching, not to vendor size. Governance and capability are separate purchases: buy the letterhead where a document must carry weight, and buy facilitation from whoever will stand in the room.
Arthur Wandzel
