AI Consulting Services

AI Consulting Services covers the work an enterprise pays for before anything gets built. Deciding what AI is actually for inside the business, what it will cost, who will own it, and how anyone will know whether it worked. It is advisory work sold on judgment rather than delivery, and for most of the last three years it has been the easiest thing in enterprise technology to sell.

That is changing, and the reason is the more interesting story in this category. The firms winning the largest engagements in 2026 are not the ones with the strongest strategic reputation. They are the ones that can carry a plan past the slide deck and into something running in production. The category is still defined as advisory, but it is increasingly being bought on engineering credibility.

What the work actually involves

Underneath the vocabulary, an engagement here answers three questions. What should we use this for. How do we run it without breaking something. How will we know it paid off.

In practice that becomes use case identification across business units, a business case and funding model, a target operating model that settles ownership, a governance and risk posture, workforce readiness, and platform selection ahead of any build. The extended version of the category picks up experience design, customer experience work, and innovation programmes, though those tend to get bought only once the first set is in place.

None of this is new consulting work. What is new is the pace at which the answers go stale, which is why the engagements have shortened and why value measurement has become a contractual item rather than a closing slide.

The line that defines the category

Data pipelines, model deployment, MLOps, lifecycle management, and everything that follows sit in a separate market called AI Technical Services. It has its own vendor evaluations and roughly eleven capabilities that fall on that side of the line rather than this one.

The distinction exists because the two markets have different buyers, different budgets, and different failure modes. It confuses people because almost every large firm sells both. A provider can be at the top of the advisory category and unremarkable at the technical one, and the handoff between the two is where a very large share of enterprise AI programmes quietly stop.

That is also the most consistent thing reference customers said when analysts interviewed them. Across every firm evaluated, the common request was continuity of team from strategy through implementation. Not better strategy. Not cheaper delivery. The same people staying on the job.

Who is in the market

At enterprise scale the recognised set is Accenture, Bain & Company, Boston Consulting Group, Capgemini, Deloitte, EY, IBM, KPMG, McKinsey & Company, and PwC.

That is the top of the market rather than the market itself. Widen the count and around thirty seven providers operate in the category once mid-market and specialist firms are included, ZS and eClerx among them. The line between the two groups is money. Making the top ten requires more than $250 million in AI services revenue over the trailing twelve months and a dedicated practice rather than an AI section bolted onto a digital transformation offering.

That threshold is worth understanding before it gets used as a quality filter, because it is not one. It answers which of the giants, not which provider. A shortlist built from the top ten only makes sense if your programme is large enough that the top ten would take the call.

What The Forrester Wave: AI Consulting Services, Q2 2026 found

The current evaluation of the category is The Forrester Wave: AI Consulting Services, Q2 2026, published in May 2026 by Ted Schadler, Sudha Maheshwari, Katie Vincent, and Ian McPherson. It ran as a multi-month assessment of those ten firms, scoring current offering, strategy, and interviews with each firm's own customers. Three findings survive summary.

The baseline is flat. Every firm in the set can do business strategy, value management, AI operations, and governance. Nobody is missing a capability. Whatever separates them sits somewhere other than the list of things they can do.

Engineering depth moved the rankings, not strategic pedigree. PwC, Accenture, EY, and IBM landed in the top tier. McKinsey and BCG, the two names a board is most likely to recognise, placed below them alongside Capgemini and KPMG. In a market where every provider can produce a credible AI strategy, the strategy stops being the differentiator and delivery becomes it.

Pricing structure turned out to carry real weight. PwC was singled out for putting fees at risk in roughly a third of its engagements. Accenture drew top scores on delivery platform and productised assets, backed by a $3 billion commitment made in 2023. EY scored highest on governance, security, and operating model, on the argument that AI moves companies away from centralised platforms toward business models built on proprietary knowledge.

One caveat on reading any of this as a leaderboard. Firms that decline to participate in an evaluation, or take part only partially, still get scored and published using secondary research. A low placement occasionally says more about how a firm manages analyst relations than about how it delivers.

Why the economics are moving

The category is being repriced by the technology it sells. Delivery platforms let firms do more work with fewer people, which puts direct pressure on the billable hour and pushes the market toward outcome-based pricing. Consultancies have historically carried almost no downside for bad advice, since the worst outcome was not being hired again. Buyers are now negotiating for genuine risk sharing, and providers are pricing for it rather than refusing it.

Two other shifts matter over the next couple of years. Engagements are moving from internal operating model work toward customer-facing and revenue-generating scenarios, which changes who signs the contract. And proprietary assets, meaning industry data, domain knowledge, and code libraries, are becoming the real separator, because that is what turns a general purpose model into something that works in a specific business.

The consulting pyramid is under pressure alongside all of this, since AI absorbs the entry-level work that used to staff it. Whatever these firms look like at the end of the decade, the shape will be different.

How to read the category as a buyer

The useful questions are narrow. Ask who carries the work from strategy into build and whether it is the same team, because that handoff is the documented failure point. Ask what proportion of fees the firm will put at risk against outcomes, because the answer is now a live negotiating position rather than an unusual request. Ask what they bring that is proprietary, because if the answer is a methodology and access to the same models you could license directly, the engagement is coordination priced as expertise.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.