AI9 August 20265 min read

AI Strategy Advisors: What They Actually Do (and When to Hire One)

A practitioner's guide to AI strategy advisors: what they deliver, what they cost, how to scope engagements, and when to build in-house instead

By AI Advisory team

Every mid-market operator has been pitched an AI strategy engagement in the last eighteen months. Most of those pitches end the same way: a 60-slide deck, a two-by-two matrix of opportunities, and a roadmap that nobody builds. The market is saturated with advisors, and the quality range is wider than in almost any other consulting category. This article is a practitioner's guide to what AI strategy advisors actually do, what separates the useful ones from the deck-generators, what a sensible engagement costs, and when you should skip the advisor entirely and hire a builder instead.

What an AI strategy advisor is supposed to do

The job, done properly, has four parts. First, understand the business well enough to know where AI can move a metric that matters. Second, audit the current operations, data estate, and tooling to identify what is achievable in the next 6-12 months versus what needs foundational work first. Third, produce a prioritised, costed roadmap that ties each initiative to a business outcome, an owner, and a realistic timeline. Fourth, help the leadership team make defensible decisions about buy-versus-build, vendor selection, and internal capability.

That's the honest version. The dishonest version - which is unfortunately common - is a generic AI opportunity landscape, a maturity model borrowed from a McKinsey publication, and a set of recommendations that could apply to any company in the sector. If the deliverable looks like it was written before the advisor met you, it probably was.

The distinction matters because the McKinsey 2024 State of AI report found that only about 15% of organisations reported meaningful bottom-line impact from generative AI investments, despite 72% adoption. The advisor's job is to get you into that 15%, not to justify the spend that lands you in the other 85%.

Types of AI strategy advisor, and what each is good for

Not all advisors are the same. The market splits roughly into five archetypes, and each solves a different problem.

The Big Four and tier-one strategy houses (Deloitte, PwC, EY, KPMG, BCG, McKinsey, Bain). Best for board-level narrative, regulatory positioning, and multi-year transformation programmes at enterprises above £500m turnover. Day rates £2,500-£5,000+. Not sensibly priced for mid-market. You get partner branding and a large team of associates who have not shipped an AI system themselves.

Boutique AI strategy firms. Typically 10-40 people, ex-consultancy leadership, focused exclusively on AI. Better sector depth than the generalists, worse operational depth than the builders. Good for a formal strategy engagement if you need external validation for a board or investors. £30k-£120k for a 6-10 week engagement.

Build-and-advise agencies (the category we sit in). Firms that do both strategy and implementation, so the roadmap is written by people who will have to build it. The advantage is honesty about feasibility - a firm that has to ship the recommendation will not put a hallucination-prone chatbot on your legal team's desk. Typical strategy engagement £15k-£40k, then implementation on top.

Independent AI advisors and fractional CAIOs. Ex-heads-of-AI who consult for 2-4 days a month across a small client roster. £1,500-£3,000 per day. Excellent for ongoing challenge and vendor triage; less useful if you need capacity to produce artefacts.

Cloud vendor professional services (AWS, Microsoft, Google Cloud). Free or heavily subsidised assessments tied to their stack. Useful for scoped technical proofs of concept on that cloud, biased toward their services. Do not confuse a vendor workshop with independent strategy.

What a good AI strategy engagement actually produces

Ignore the format of the deliverable. Judge the engagement by what the leadership team can do on the Monday after it ends. A good AI strategy advisor leaves you with:

A prioritised opportunity map, ideally 12-25 candidate initiatives, each scored on business value, feasibility with current data and tooling, and estimated cost. Not a two-by-two - a table with numbers.
A data and tooling audit that names the specific systems, integrations, and data quality issues that will constrain execution. If the advisor has not connected to your CRM, warehouse, or ticketing system, they have not audited anything.
A 12-month roadmap broken into quarters, with realistic assumptions about internal capacity. If it assumes you will hire three data engineers in Q2, the advisor is dreaming.
Vendor and platform recommendations with defensible reasoning. "Use OpenAI" is not a recommendation. "Use Azure OpenAI with a private endpoint because your data is in Microsoft 365 and your DPO will not sign off on a US processor without EU data residency" is a recommendation.
A governance framework proportionate to your risk. Under the EU AI Act, which entered into force on 1 August 2024, obligations for general-purpose AI models applied from 2 August 2025, and high-risk system obligations phase in through 2026-2027. UK-only businesses still fall under ICO guidance on AI and data protection. A strategy engagement that ignores this is incomplete.
Business cases for the top three initiatives, detailed enough to hand to a build team without further discovery. This is the artefact most advisors skip because it requires real work.

If the final deliverable is a deck without an accompanying business-case document per initiative, you have paid for a workshop, not a strategy.

What it costs, and how to scope the engagement

Sensible price ranges for mid-market UK businesses in 2026:

Two-week readiness assessment: £8k-£20k. Enough to identify the top 3-5 opportunities, do a light-touch data audit, and produce a costed 90-day plan. Right size for companies that just need to get moving.
Six to eight-week full strategy engagement: £30k-£80k. Full opportunity mapping, deeper data audit, governance framework, 12-month roadmap, business cases for top initiatives.
Ongoing fractional advisory: £4k-£15k per month. Regular executive challenge, vendor triage, review of build outputs.

Anchor your budget to what happens after the strategy. Productive's 2024 agency benchmark data puts average project margins in the professional services category at 20-30%; a strategy deliverable that does not lead to implementation is pure cost. If you cannot fund at least the first initiative on the roadmap, do not commission the strategy yet - do a two-week readiness assessment instead and use the output to unlock the build budget.

Contract structure matters. Prefer fixed fee over time-and-materials for strategy work; the scope is knowable and open-ended engagements drift. Insist on named individuals in the statement of work - AI strategy quality collapses when the pitch team hands off to junior associates.

How to evaluate an advisor before you sign

Six questions that separate operators from deck-generators:

1. Show me the last three roadmaps you produced, redacted. Not case study one-pagers - the actual artefacts. If they cannot show you comparable output, they have not done comparable work.

2. Which of your recommendations from the last twelve months got built, and by whom? An advisor with a low implementation rate is producing shelf-ware. Either they cannot write buildable recommendations, or they cannot influence clients to act on them. Both are disqualifying.

3. Who on the pitch team will be on the delivery team? Get names, allocations, and a clause in the SOW that stops substitution without your consent.

4. Walk me through a recommendation you made that failed. If they cannot name one, they are either new to the work or lying. Both are problems.

5. What is your position on open source versus proprietary models for our use cases? A useful advisor has a defensible opinion informed by your specifics - data sensitivity, latency requirements, budget, in-house skill. "It depends" without follow-up is not an answer.

6. How do you handle the EU AI Act and ICO guidance in your recommendations? If the answer is vague, they will produce a strategy your DPO rejects. The ICO's guidance on AI and data protection is not optional reading for anyone advising UK companies.

When you should skip the advisor entirely

Not every AI initiative needs a strategy engagement. Skip the advisor and go straight to a builder when:

The use case is already obvious and bounded. If your support team is drowning in tickets and you know a RAG-grounded assistant on your knowledge base will help, you do not need a six-week strategy. You need a scoped build.
The budget is under £30k total. A meaningful strategy engagement will consume most of it and leave nothing to build. Better to run a discovery-plus-build sprint where the discovery is 1-2 weeks and directly informs a working prototype.
You have internal AI capability already. If you have a head of data or head of engineering who has shipped ML or LLM systems, they can do the opportunity mapping. Bring in fractional advisory for challenge, not a full engagement.
You are pre-product-market-fit. AI strategy for an unproven business is premature optimisation. Fix the product first.

Conversely, commission a proper strategy engagement when the AI investment is material (>£150k across the year), when there are cross-functional dependencies you cannot resolve internally, when regulatory exposure is significant (financial services, healthcare, legal), or when the board needs an external artefact to release budget.

The build-and-advise model, and why it changes the incentives

The traditional consulting model separates strategy from execution. The strategist writes the recommendation and leaves; someone else - often a systems integrator on a much larger contract - builds it. This creates two well-known failure modes. The strategist optimises for the impressiveness of the recommendation rather than its buildability, and the integrator optimises for scope expansion rather than the intent of the strategy.

Firms that do both under one roof have different incentives. If you write a roadmap you will have to build, you write a buildable roadmap. If you build against a roadmap you wrote, you cannot blame the strategist for the constraints. The trade-off is that these firms are usually smaller and lack the brand weight of a tier-one strategy house. For most mid-market engagements that trade is worth taking - you get honesty about feasibility and continuity between the plan and the ship.

The signal to look for: does the firm publish its build stack, its evaluation methodology, and the specific tools it uses? A pure strategy firm will be vague about implementation because they do not do it. A builder-advisor will name n8n, LangChain, pgvector, Azure OpenAI, and explain why each is chosen for which problem. If the firm cannot tell you what they build with, they do not build.

Frequently asked questions

How long does a typical AI strategy engagement take?

A two-week readiness assessment covers a light audit, top opportunities, and a 90-day plan - suitable for teams that need to move quickly. A full strategy engagement typically runs 6-8 weeks: two weeks of discovery and stakeholder interviews, two weeks of data and tooling audit, two weeks of opportunity modelling and business case development, and a final week to socialise the output with the executive team. Anything longer than ten weeks usually means the advisor is padding, or the scope has drifted into implementation planning that should be priced separately. Beware of engagements that stretch to twelve weeks or more without a clear reason.

What should an AI strategy roadmap actually contain?

At minimum: a prioritised list of 12-25 opportunities scored on value, feasibility and cost; a data and tooling audit naming the specific systems reviewed; a 12-month quarter-by-quarter plan tied to internal capacity; vendor and platform recommendations with defensible reasoning; a governance framework proportionate to your risk profile including EU AI Act and ICO exposure; and detailed business cases for the top three initiatives. If the deliverable is only a deck, you have paid for a workshop. Insist on the underlying artefacts - the scored opportunity register and the business cases are what let your team act on the strategy.

How much should a mid-market business budget for AI strategy?

For UK mid-market businesses (50-1,000 employees), expect £8k-£20k for a two-week readiness assessment or £30k-£80k for a full 6-8 week engagement. Fractional ongoing advisory runs £4k-£15k per month depending on the seniority and time commitment. As a rule of thumb, the strategy should be no more than 15-20% of the first-year AI budget - if it consumes more, you will not have enough left to build anything meaningful. If your total AI budget is under £30k, skip the standalone strategy engagement and commission a combined discovery-and-build sprint instead where the planning directly produces a working system.

Can we do AI strategy in-house instead of hiring an advisor?

Yes, if you have the right people. A head of data or head of engineering who has shipped ML or LLM systems can lead opportunity mapping, and your operations leaders know where the pain is. What in-house teams typically lack is exposure to what other companies have tried and failed at, and the discipline to produce a formal roadmap alongside their day job. A hybrid model works well: run the strategy in-house and bring in fractional advisory (2-4 days a month) for challenge, vendor triage, and to prevent groupthink. Full external engagements are most useful when there is no internal AI leadership yet or when the board requires an independent artefact.

What is the difference between an AI strategy advisor and an AI consultant?

The terms are used loosely, but the useful distinction is scope. A strategy advisor focuses on the executive-level questions: which problems to solve, in what order, with what governance, and at what cost. An AI consultant typically operates one layer down - helping design a specific solution, selecting models, defining evaluation criteria, or supporting a build team. You need both, but not always at the same time. Start with strategy when the question is "where should we invest?" and move to consulting or build when the question is "how do we implement this specific initiative well?" Many firms offer both, which reduces handoff friction.

How do we make sure the AI strategy actually gets implemented?

Three tactics work. First, only commission a strategy when you have the budget and executive sponsorship to fund at least the first initiative on the roadmap - otherwise you are producing shelf-ware. Second, insist that business cases for the top three initiatives are part of the deliverable, detailed enough to hand to a build team without further discovery. Third, prefer advisors who also build, or line up an implementation partner in parallel with the strategy so there is no gap between finishing the plan and starting the work. The riskiest handoff in AI programmes is between the strategist walking out and the builder walking in - close it.

How do we evaluate an AI strategy advisor's track record?

Case studies are marketing artefacts, not evidence. Ask for three concrete data points: what percentage of their recommendations from the last twelve months were implemented, by whom, and with what measured outcome. Ask to speak to two references whose engagements are more than nine months old - long enough for the strategy to have been tested against reality. Ask them to walk you through a recommendation that failed and what they learned. Advisors who cannot answer these questions concretely are producing decks, not strategies. A strong track record shows implementation rates above 60% and references who can name specific business outcomes, not just "good workshop".

Does the EU AI Act affect UK-only businesses?

Directly, only if you place AI systems on the EU market, provide AI services to EU users, or the output of your AI is used in the EU. Many UK mid-market businesses meet at least one of those tests. UK-only businesses are governed by UK GDPR and the ICO's guidance on AI and data protection, plus sector regulators (FCA for financial services, MHRA for medical devices). A useful strategy engagement maps your specific exposure - it is not enough for an advisor to say "we'll follow best practice". Ask them to name the specific obligations that apply to your intended use cases and how the roadmap accounts for them.

Choosing well matters more than choosing quickly

The AI strategy market will remain crowded and uneven for the next few years. The advisors who will be worth hiring in 2027 are the ones producing roadmaps in 2026 that clients actually build against. Evaluate them on their artefacts, their implementation rates, and their willingness to be specific about tools, costs, and constraints. If the pitch is generic, the strategy will be generic. If the pitch names the systems in your stack and the regulators you answer to, you are talking to someone who does the work.

At AI Advisory we run strategy engagements as the front end of a build relationship, because we think the honest test of a recommendation is whether it ships. If you want to talk through where an advisor fits into your plans - or whether you should skip straight to a build - get in touch.