How much does AI consulting cost in 2026? Day rates, fixed prices and what drives the bill

What the market actually quotes

The most checkable price signal in this market is not a consultancy's rate card. It is what employers write in job ads when they need the same skills for a few months. IT Jobs Watch reads those ads and publishes the distribution.

Read that as a floor for firm pricing, not as the price of a firm. A consultancy quoting a day rate is also selling delivery management, cover when someone is ill, and the risk that the work takes longer than planned. What it should not be selling is the privilege of a logo.

Why most published price ranges are useless

Search for the cost of AI consulting and you will find tidy tables: solo experts at one band, boutiques at another, large firms at a third. Almost none of them say who was surveyed, how many, or when. They are written by the same consultancies whose prices they describe.

Three questions separate a number you can use from a number someone made up. Who was counted? How many? On what date? A range that cannot answer all three belongs in a pitch deck, not in your budget.

This is why the table above cites a sample size and a date, and why the only price we quote for our own work is the one on our price list.

Fixed price or time and materials

Both are honest. They answer different questions.

• Fixed price fits a decision. An assessment, an audit, a roadmap: the work has a defined end and a written deliverable, so the price can be written down before it starts. You can put two offers side by side and compare them.
• Time and materials fits a build. Modelling on data nobody has profiled yet changes shape as you learn. A fixed price here is either padded for risk or renegotiated later, and both outcomes cost you more than paying for the days you use.

The practical rule: fix the price of the first step, then decide about the build once you know what the data allows. A vendor who refuses to fix the price of a two-week assessment is telling you something about how they scope work.

What a readiness audit costs and what it contains

Our own number, because we publish it: 1 400 EUR net, ten working days from the first interview, one payment, no retainer and no obligation to buy anything afterwards.

The sequence is fixed: a 90-minute interview with the person who owns the decision, up to three 45-minute interviews with the people who actually run the processes, five working days of analysis, then a 60-minute session to walk through the report.

The report contains a process map with an automation score, an assessment of whether your data can carry the use cases you have in mind, an impact-versus-effort matrix, three recommendations with cost and time estimates, the competence gaps in your team, and the list of things we would not build in your situation and why.

That last section is the one clients quote back to us. It is also the reason the price can be fixed: we are selling a decision, not a build.

What really drives the bill

1. The state of your data. Gartner expects most organisations to abandon AI projects that are not supported by AI-ready data. Profiling, cleaning and joining data is where the hours go, long before anyone trains a model.
2. How many people have to agree. Each additional decision maker adds workshops, revisions and waiting time. Two people who can say yes will always be cheaper than six who can only say maybe.
3. The size of the first step. A month of discovery before any code is a cost that produces slides. A two-week assessment that ends in a decision is a cost that produces a direction.
4. Compliance context. Regulated processes need documentation, traceability and review, and that is real work, not padding.
5. Who does the work. A named senior person on the account costs more per day and less per project than a rotating team you meet after signing.

How much budget is normal

BCG's 2026 survey of 2,360 executives across 16 markets found companies planning to spend an average of 1.7 % of revenue on AI, more than double the 0.8 % of 2025. Financial institutions and technology companies sit near 2 %, industrial and real estate companies below 1 %.

Use it as a sanity check, not a target. Our own arithmetic on the BCG figure: for a company with 20 million EUR in revenue, 1.7 % is 340 000 EUR a year across tools, people and outside help. Against that, a fixed-price assessment is a rounding error, and it decides where the rest goes.

Frequently asked questions

Antonina

Research Analyst, Akademia Data Science