Which AI Consulting Company Should You Choose?

No Big Four firm publishes a price for building you an AI tool. Not one. Deloitte, PwC, EY and KPMG have announced roughly $7.4 billion in AI investment between them. None of it has produced a rate card. No starting price either, and no published example of what a finished build costs.

So we went to the one place their AI work is a matter of public record: US federal contracting. Every prime contract award to all four firms is published, with the dollar amount and a description of the work.

What's there is not what the announcements suggest.

Key Takeaways

• Across 1,607 federal prime contract awards to the four firms active in FY2025 — $21.3 billion obligated — six name AI in the description. That's 0.03% of the dollars.
• Those six total $5.9 million. The median is $265,245. The largest is $3.9 million.
• Not one of the six is a tool build. They're program management, forecasting, a talent-gap study, maintenance of an existing application, and two reviews.
• The best-documented Big Four AI deliverable anywhere is an AUD $439,142 report for the Australian government. It came with an AUD $97,587 refund after AI-fabricated citations turned up in it.
• Announced AI investment: Deloitte over $3bn through FY2030, KPMG $2bn, EY $1.4bn, PwC $1bn. The gap between that and the contracted work is the thing to notice.

Table of Contents

• What Does Big Four AI Tool Development Cost?
• Why None of Them Publishes a Price
• The $7.4 Billion They've Announced
• What They Actually Get Paid For
• Six Awards, and Not One Is a Tool Build
• The One AI Deliverable With an Exact Price
• Is Tool Development Even What They Sell?
• What This Means at 30 People
• How to Price-Check a Big Four AI Proposal
• Your Next Step
• Frequently Asked Questions
• Sources
• About Avolis Research Group
• Continue Learning

What Does Big Four AI Tool Development Cost?

There's no published number, and the absence is the answer. US federal contracting is the only public record of what these firms are paid for AI work. In fiscal 2025, six awards name AI. They're worth $5.9 million combined, against $21.3 billion in total obligations to the four firms. Median: $265,245.

Read that median carefully, because it's not a build price. It's the price of six engagements that are mostly advisory, and we'll walk through each one below. The useful finding isn't a dollar figure you can budget against. It's that the thing you're shopping for barely shows up in what these firms contract to do. A firm that builds an AI tool and hands it over. That category.

That matters more than a rate card would. A number would tell you what something costs. What the record tells you instead is whether you've walked into the right shop at all — the more expensive question to get wrong.

Why None of Them Publishes a Price

Partly because nothing makes them. Big Four AI work is sold as a scoped engagement, negotiated per client, and priced off a blended rate the client never sees broken out. There's no list price because there's no list product.

But there's a second reason, and it's the one worth your attention. What gets sold under "AI" at this tier is usually not a discrete tool with a fixed scope and a fixed deliverable. It's an assessment, a strategy, a governance framework, a program office, or a platform rollout. Work whose size depends entirely on how big your organization is. A price would be meaningless. Ours is $X assumes a standard unit of work, and there isn't one.

You'll find plenty of pages online quoting Big Four AI rates of $300 to $1,000 an hour. We checked. Almost none of them cite a source, and the ones that do cite each other. Treat those numbers as fiction until someone shows you the document. Our piece on alternatives to big AI consulting firms works through the only day rates these firms publish anywhere: UK government framework listings. That's a different record from this one.

The $7.4 Billion They've Announced

All four firms have made large, public AI investment commitments, and they're easy to verify because each firm announced its own. Deloitte has committed over US$3 billion in generative AI through FY2030. KPMG pledged $2 billion over five years. EY put US$1.4 billion behind the EY.ai platform. PwC US committed $1 billion over three years.

Note what these numbers are. They're each firm's investment in its own capability — platforms, licences, training, acquisitions. They are not what a client pays, and they're not evidence that the money produced sellable tool-development capacity. Deloitte's release, for instance, sits inside its FY2025 results, where the firm reported US$70.5 billion in revenue and over 470,000 people. The $3 billion is a commitment through FY2030 against that base.

Announced investment is a marketing number as much as a financial one. It tells you a firm wants the AI category. What it doesn't tell you is what that firm delivers to a client, what delivery costs, or whether any of the money bought capacity you could buy time on.

What They Actually Get Paid For

The US government publishes every prime contract award: recipient, dollar amount, agency, and a description of the work. That makes federal contracting the largest checkable record of Big Four delivery anywhere. So we pulled all of it.

What we did, so you can check it: on 19 August 2026 we queried the USAspending.gov API (/api/v2/search/spending_by_award/) for every prime contract award to Deloitte, PricewaterhouseCoopers, KPMG and Ernst & Young. Award types A, B, C and D. The window: 1 October 2024 to 30 September 2025, US fiscal year 2025. That returned 1,607 awards totalling $21,317,500,902 in obligations. We then searched each award's description for AI language: "artificial intelligence," "machine learning," "generative AI," "large language," "GenAI," "AI/ML," or "AI" as a standalone word. Six awards matched, worth $5,905,673 combined. That's 0.0277% of the dollars and 0.373% of the awards. The API is public and takes no key; anyone can reproduce this in about ten minutes.

Two caveats before you use those figures, because they change what the numbers mean. Each award's amount is its total obligation to date, not FY2025-only spending — several of these awards began years earlier and were still active. And the description field is written by the contracting officer, so an award can involve AI work without saying so. This is a floor on AI-labelled work, not a ceiling on AI work.

Even as a floor, the shape is stark. The firms are not evenly represented either. Deloitte alone accounts for $16.79 billion across 1,249 awards, which is 78.8% of every federal dollar the four firms took. EY follows at $2.67 billion, KPMG at $1.84 billion. PwC has nine awards worth $17.0 million. Eight-hundredths of one percent. "The Big Four" is a category with very different businesses inside it.

Six Awards, and Not One Is a Tool Build

Here's every federal award to the four firms that names AI, in full. Read the descriptions rather than the numbers, because the descriptions are the finding.

Now line the six up against the phrase you searched for. One is program-management support for someone else's AI program. One is forecasting for a recruiting office. One is a study of what AI skills the Air Force needs to hire. One maintains an application that already exists. Two are reviews — a root-cause analysis, and an assessment of whether the Occupational Safety and Health Administration should be using AI at all.

None of them is "build us an AI tool."

That's six for six, and the pattern holds when you widen the net. Loosen the search to analytics, automation, data science, robotics and modelling — the language these firms actually use. That gets you 62 awards worth $946 million, 4.4% of the dollars, at a median of $4.1 million. Bigger work, and closer to systems. Still overwhelmingly staffing and analytical support rather than a firm building software and handing it over.

The One AI Deliverable With an Exact Price

Want a single Big Four AI engagement with a published price, scope, and outcome? There's one. It isn't in the US. Australia's government contract register carries the whole thing.

Deloitte Touche Tohmatsu was engaged by Australia's Department of Employment and Workplace Relations for an assurance review supporting the Targeted Compliance Framework. Contract value: AUD $439,142.00. Period: 23 December 2024 to 4 July 2025. Open tender, five suppliers invited. The deliverable was a 237-page report.

After publication, a University of Sydney researcher found citations in it to academic work that did not exist. Also a fabricated quote from a federal court judgment. Deloitte confirmed the errors and issued a revised version that disclosed the use of Azure OpenAI in producing it. The government's own contract notice now carries the resolution in its description field: "Final Instalment of $97,587.11 (GST inclusive) repaid to department."

We're not raising this to score a point about one firm. Every organization using these tools, ours included, has to build checks that catch this. We raise it because it's the most completely documented AI-assisted Big Four deliverable in existence. Price, scope, method, defect, remedy, all on a government domain. And what it documents is a report. Not a tool. A document that needed a refund.

Is Tool Development Even What They Sell?

Mostly, no. Read the four firms' own AI announcements and the product is a platform they operate, not software they build for you. Deloitte has Zora AI. PwC has agent OS. EY has EY.ai, which the firm describes as "a unifying platform that brings together human capabilities and artificial intelligence." KPMG has Workbench. These are the firms' own systems, sold as access plus the consultants who configure them.

That's a genuinely different purchase from tool development, and the difference decides whether you should be talking to them at all. In a platform engagement you pay for licence, configuration, change management and advisory. You don't end up owning a system built around how your business works. You end up as a tenant on theirs.

For a large enterprise that's often the right trade. Standardization has real value at 40,000 employees. At 40 employees it usually inverts. A workflow drains your team precisely because it's specific to your operation. Your intake. Your estimating quirks. Your scheduling constraints. A platform built to standardize across thousands of clients is the wrong shape for that by construction.

What This Means at 30 People

Practically, three things. First, the number you can't find isn't being hidden from you — it doesn't exist, because the product you're imagining isn't the product on offer. Second, the median AI-named engagement these firms take from the US government is $265,245, and that buys advisory work, not a system. Third, and most usefully: if this is what the federal government gets, a 30-person operation isn't going to get something categorically better for less.

There's a version of this that sounds like we're just knocking the competition. So let's be concrete about when the answer flips. Need an auditable governance framework because a regulator is asking? Or a board-defensible assessment before a large capital commitment? A Big Four name carries weight in those rooms that ours simply doesn't, and pretending otherwise would cost you more than the fee does. Buy the deck. It's doing a real job.

But if you want the estimating backlog to stop eating fourteen hours a week, that's a build. And a build wants someone in your operation who owns the outcome. Someone who stays. Our piece on the best AI consulting firms for small businesses works through which type of partner fits which sector. The pillar on choosing an AI consulting company covers the evaluation itself.

How to Price-Check a Big Four AI Proposal

If you have a proposal in hand, the two records in this article give you something a general negotiation guide can't. Use them in this order.

Ask which of the four things you're buying. Licence to their platform, configuration of it, advisory hours, or custom development. Make them split the quote by those four lines, itemized, because most proposals blend all four into a single number and the blend is exactly where the margin hides.

Then ask for a comparable. Not a case study — a comparable, meaning a named engagement of similar scope with the price. If they can't produce one, that's information, and it's the same information the public record already gave you.

Then check the record yourself. Search their name on USAspending.gov and read what they've actually been contracted to do. It's free, it takes ten minutes, and it's the same firm bidding on your work.

Finally, convert everything to a twelve-month total, including your own people's time. All of it. A quote that looks like $180,000 of build is often $180,000 of build plus a platform licence that renews every year after that. Our guide to how much AI consultants charge covers the pricing models across the whole field.

Your Next Step

Don't start by choosing a firm. Start by knowing which workflow is actually costing you, and what a fix would be worth. That number, not the vendor's, decides what you should pay and who's the right size to call.

That's what a diagnostic is for. Want a picture of where AI would pay off before anyone quotes you anything? An AI readiness assessment is the place to start. And if the honest answer is that there's nothing worth automating yet, we'll tell you that instead of selling you a project.

Frequently Asked Questions

What do Big Four consulting firms charge for AI tool development in 2025?

None of the four publishes a price, and no credible source establishes one. In US federal contracting — the largest public record of their work — six awards active in FY2025 name AI, totalling $5,905,673, with a median of $265,245. All six are advisory, program-management or maintenance work rather than tool builds.

Which of the Big Four does the most AI work?

By federal dollars, Deloitte is far ahead: $16.79 billion across 1,249 awards active in FY2025, or 78.8% of the four firms' total. But that's overall work, not AI work. Of the six AI-named awards, four are Deloitte's and two are KPMG's. PwC and EY have none.

Are the $300 to $1,000 per hour Big Four AI rates quoted online real?

Treat them as unverified. We could not find a primary source for any of them, and the pages quoting them largely cite one another rather than a firm's own document. The only day rates these firms publish anywhere are on UK government procurement frameworks, which are public-sector prices in pounds.

Why is PwC nearly absent from US federal contracting?

PwC held nine prime contract awards worth $17.0 million active in FY2025 — 0.08% of the four firms' federal total, which is close to absence. We're reporting the figure, not the cause; the register records awards, not the commercial reasons behind them. But "Big Four" clearly describes four different businesses.

Should a small business ever hire a Big Four firm for AI?

Sometimes. If you need a governance framework a regulator will accept, or an assessment a board will defend, the brand does real work. If you need a workflow rebuilt so it stops costing you hours every week, you're buying a build, and these firms mostly sell platforms and advisory instead.

Is announced AI investment a good signal of capability?

Not on its own. The four firms have announced roughly $7.4 billion, and that spending is real. But it's investment in their own platforms, licences and training — not evidence of what they deliver to clients, and not something you can price a project against.

Sources

All sources retrieved 2026-08-19 unless noted.

• USAspending.gov, Award Search API (/api/v2/search/spending_by_award/) — the 1,607 awards, $21,317,500,902 in obligations, the per-firm splits, and the six AI-named awards are our own query and arithmetic, not a published USAspending figure. Prime contract awards only (types A–D), recipient search on Deloitte, PricewaterhouseCoopers, KPMG and Ernst & Young, contracting activity 1 Oct 2024 – 30 Sep 2025. Amounts are total obligations to date on each award, not FY2025-only spending; description text is written by the contracting officer, so AI-labelled work is a floor rather than a full count.
• AusTender, Contract Notice CN4118426, Department of Employment and Workplace Relations / Deloitte Touche Tohmatsu — "Assurance Review to support the Targeted Compliance Framework. Final Instalment of $97,587.11 (GST inclusive) repaid to department." Contract value AUD $439,142.00; period 23 Dec 2024 – 4 Jul 2025; open tender, five suppliers invited; consultancy, "need for independent research or assessment." Published by the Australian Department of Finance under CC BY 3.0 AU.
• Deloitte Global, Deloitte reports FY2025 revenue, 30 September 2025 — US$70.5 billion revenue, 4.8% growth in local currency, over 470,000 people, and "over US$3 billion of investments in GenAI through FY2030."
• EY, EY announces launch of artificial intelligence platform EY.ai following US$1.4b investment, 13 September 2023 — the US$1.4 billion figure, covering roughly the preceding 18 months, and the platform description quoted above. No client pricing appears in the release.
• PwC US, PwC US makes $1 billion investment to expand and scale AI capabilities, April 2023 — $1 billion over three years, in an expanded relationship with Microsoft and OpenAI. PwC US only, not global.

• CFO Dive, "KPMG to spend $2B in expanded AI alliance with Microsoft," 11 July 2023 — $2 billion in Microsoft cloud and AI services over five years. Trade-press reporting; we could not locate a corresponding release on KPMG's own newsroom, so the exact $2 billion figure rests on secondary coverage while the other three rest on the firms' own releases. KPMG's own Workbench release, cited below, does describe "a multi-billion dollar investment in AI and agentic transformation" without giving a number.

• The four platforms named above, each from the firm's own announcement: Deloitte, Introducing Zora AI (launched at NVIDIA GTC, March 2025); PwC US, PwC launches AI agent operating system to revolutionize AI workflows for enterprises (agent OS, 2025); EY.ai (September 2023); and KPMG International, KPMG launches a multi-agent AI platform (KPMG Workbench, 17 June 2025, built on Microsoft Azure AI Foundry). All four are platforms the firms operate, which is the point made in the section above.

• Fortune, "Deloitte to refund Australian government for report with AI hallucinations," 7 October 2025 — the 237-page length, the Azure OpenAI disclosure in the revised report, the fabricated judgment quote, and University of Sydney researcher Chris Rudge's identification of the errors. Note that press coverage of this contract has circulated several different dollar figures; the AUD $439,142.00 value and AUD $97,587.11 refund above are taken from the Australian government's own contract notice, which is authoritative.

About Avolis Research Group

Avolis Research Group is Avolis's in-house research practice, focused on how operations-heavy small and mid-sized businesses actually adopt AI. It synthesizes primary economic research, government survey data, and results from real implementations into practical, vendor-neutral guidance.

More about Avolis and how we work · Get in touch

Continue Learning

• Which AI consulting company should I choose? — the pillar: how to run the whole decision.
• What are good alternatives to big AI consulting firms? — the published UK day rates, and what replaces which part of the big-firm bundle.
• Best AI consulting firms for small businesses — which type of partner fits which sector.
• How much do AI consultants charge? — pricing models across the whole field.
• AI readiness assessment services — diagnose where AI pays off before anyone quotes you.