Guide
AI consulting has become a label for everything from a two hour workshop to a £20m programme. Here is what the work really involves, what you should pay for it, and the questions that separate consultants who have delivered from consultants who have slides.
By Christopher Fagan, Managing Partner, PEF Consulting Partners
There are four distinct jobs inside the phrase. An AI readiness assessment looks at your data, skills, governance and candidate use cases and tells you honestly what is possible now. An AI strategy decides where AI will pay in your business, in what order and with what budget. AI governance builds the framework, controls and compliance evidence that let you use it safely. Implementation is the building or integrating of the thing itself, including vendor selection, data work, testing and the change in how people work. Most organisations need the first job and one or two real use cases before they need a strategy document, and the consultancies that insist on the strategy first are usually the ones paid by the slide.
The large consultancies bring scale and alliances with Microsoft, Google and Amazon, at the highest rates and with a strong pull towards whichever platform they are partnered with. The vendors and their partners sell a platform and consult around it, which is fine once you have chosen it. AI agencies and boutiques build quickly and well, and are often weak on governance and on integration with the core systems that hold your data. Independent consultants work client side, with no product to sell, and are the right choice for the assessment, the strategy, the governance and for holding the builders to account. Few organisations need only one of these, and the client side role is the one most often missing.
Day rates for experienced independent AI consultants run from around £800 to £1,800, and from £1,500 to well over £3,000 for senior people at the large firms. A readiness assessment is a fixed scope job and should be a low five figure sum. Taking a first use case into production typically costs between £50,000 and £250,000 depending on the data work and the systems it has to connect to. Full programmes run from six figures to tens of millions. The fifteen or so AI programmes we have delivered, from government to a hedge fund, ranged from £100,000 to £27m, and the smaller ones were not the less valuable ones.
Pilots that never leave the lab because nobody owned the path to production. Data that turned out to be incomplete, inconsistent or unlawful to use once someone looked. Governance bolted on after a problem rather than designed in. A vendor platform chosen before the use case was understood. And measuring activity, such as the number of pilots or users, instead of value, such as hours saved, errors removed or revenue won. Every one of these is a management failure rather than a technical one, which is why the consultant you choose matters more than the model.
Ask to see something they have put into production, and who exactly did the work. Ask what it cost and what it saved, with the client willing to say so. Ask what they sell, including partner commissions and referral fees. Ask how your data will be handled, where it goes and whether it trains anyone else’s model. Ask what happens when the system is wrong, because it will be. And ask how your own people will run it after the consultants leave. The answers to those six questions are worth more than any credential.
PEF does the client side jobs: a fixed scope AI readiness review, a strategy you can fund, a governance framework your board and auditors can live with and, where it is wanted, the delivery of the use cases themselves. We have no tools to sell and no platform partnerships, which is what lets us tell you when the right answer is not to build anything yet.
Related reading: AI consultants in London, AI readiness, governance and implementation consultancy, the free AI readiness check, a practical AI governance framework, where to start with AI when your business feels lost.
An agency builds things: models, assistants, automations. A consultant decides what should be built, whether it is safe and worth it, and holds the builders to account. The best outcomes usually involve both, with the consultant on the client’s side of the table.
Not a long one. You need a clear view of your data, a short list of use cases that pay back within a year and a governance baseline. A full strategy makes sense once one or two use cases are live and you understand what AI actually does in your business rather than in a vendor demonstration.
Three to six weeks for a mid sized organisation. It covers data, skills, governance, systems and candidate use cases, and ends with a written view of what is possible now, what needs work first and what is not worth doing. Our free twelve question self check gives a first indication in a few minutes.
Yes, if it is scoped properly. An SME does not need a strategy programme; it needs someone to find the two or three places where AI will save real money, check the data and the risks, and get the first one working. That is weeks of work, not months, and it should be priced as such.
If you are comparing AI consultants and want an honest view of what you actually need, a short confidential conversation costs nothing.