An AI implementation consultant earns the day rate when your blocker is sequencing work across teams that do not report to each other. If the blocker is that nobody on the payroll can build a retrieval pipeline or put a model behind a latency target, you are buying the wrong role, and you will find out around week six when the decision log is full and the repository is empty.
The title covers at least three different jobs. Buyers who do not separate them before going to market pay architect rates for meeting facilitation.
Three jobs sold under one title
| What they sell | What lands in your repo | Indicative day rate |
|---|---|---|
| Strategy advice, relabelled as implementation | Roadmap, use-case shortlist, business case | €1,200–€2,500 |
| Delivery architecture plus hands on keyboard | Reference architecture, first production path, code review | €900–€1,600 |
| Programme management for an AI portfolio | RAID log, steering material, vendor coordination | €700–€1,200 |
Rates move with country, seniority and whether the person sits inside a practice with a bench to carry. Our published figures for engineers sit on the daily rates page; a consultant with the same CV and a partner above them costs 40% to 90% more for the same hours.
The middle row is what most buyers actually want, and it is the hardest to verify from a CV. The strategy advisor and the delivery architect describe their last engagement in nearly identical language.
What moves the number
Only one of the three price mechanisms is about engineering skill.
- Shipping history. People who have taken an LLM or forecasting system through security review, load testing and a year of drift are rare, and that scarcity is worth roughly 30% to 50% over a competent engineer who has only built pilots.
- Who carries the estimate. A consultant on time and materials prices lower than the same person inside a fixed-scope engagement, because the fixed price contains a contingency you cannot see. We set out how each category of firm builds that contingency in our breakdown of how AI consulting firms price work.
- Overhead above the biller. Practice leadership, sales cost and utilisation targets all sit in the rate. A two-person boutique and a global practice can field the same engineer at rates that differ by a factor of two.
Four questions before you sign
- Ask for the last three production systems they touched, with the integration that caused the most pain in each. Vague answers mean strategy work.
- Ask who writes the first 500 lines. If the named consultant does not, find out who does and interview that person instead.
- Ask what they hand over at the end and who maintains it in month seven. Architecture documents without an owner decay in a quarter.
- Ask for the ratio of advisory days to build days in the proposal. Above 30% advisory, you are funding a discovery habit.
Where the budget leaks
Discovery is the common leak. A six-week discovery phase on a project with a six-month build consumes 20% to 25% of the budget before a single component exists, and much of it re-documents what your own team already knows. Two weeks is usually enough when the consultant can read code.
The second leak is the handover that never happens. If nobody in-house reviewed the pull requests, you have rented a system rather than acquired one, and the renewal conversation starts from a weak position.
When our model is the wrong one
We rent senior AI and data engineers on time and materials, so read this with the obvious caveat attached. The model fails in three situations.
If you have no in-house technical owner, rented engineers will build what they think you meant, and you will have no way to tell whether that was right. If you need somebody accountable for a business outcome rather than for a system, a consultant with a scoped mandate fits better than hourly capacity. And if the work is a two-week audit with a written opinion at the end, a day-rate engineering contract is clumsy. Buy the opinion.
Where the model does fit, the shape is simple: a named technical owner on your side, two to four senior engineers from ours, and a review cadence you control. That is the arrangement behind our AI consulting work, and the people doing it are the same AI engineers you would hire directly, billed by the day rather than placed.