feed

AI strategy consulting: what a good engagement should deliver (and how to tell it will ship)

Lauren
Lauren Head of Growth · 16 Nov 2026 · 12 min read

AI strategy consulting is outside help to decide where AI should change how your business works, what to build or buy, and in what order. A good engagement hands over more than a deck. You should leave with a map of the workflows worth changing, a build or buy call for each, a proof of concept on a sample of your own data, agreed success measures, and a plan your own team can run.

This guide is for heads of operations, product and digital leaders at UK scale-ups and enterprises who are about to hire an AI strategy consultant. It covers what the work should produce, the signs that a strategy will never ship, how to choose the kind of firm, and the questions to ask before you sign. If you are still deciding whether you need a consultancy at all, start with how to choose an AI consultancy.

What is AI strategy consulting?

An AI strategy consultant looks at how your business runs, finds the places where AI would change a result you care about, and helps you decide what to do about each one. The output is a set of decisions: which workflows to change, whether to build, buy or wait, what to test first, and who owns the result.

An AI strategy workshop with your leadership team is often the first step. The people who own the workflows have to agree the problem before anyone can solve it, and a workshop puts them in one room.

Good AI strategy is specific. It names the workflow, the data it runs on, the people who use it and the measure that should move. A strategy that talks about AI in general terms, with no workflow named, gives your team nothing to build.

What should a good AI strategy engagement hand over?

By the end, you should hold five things. Your team can act on each one without the consultant in the room.

  • A workflow map. The processes your business runs every week, where the time and money go in each, and which ones AI could change. The map ranks them, so you know where to start.
  • A build or buy call. For each workflow worth changing, a decision to build something you own, buy a tool, or leave it alone for now, with the reason. Some workflows are better served by a tool you can license today.
  • A proof of concept. A working prototype of the top-ranked idea, tested on a sample of your own data, that shows whether the idea works before you fund the build. A prototype on made-up data proves only that the demo works.
  • Success measures. The number each change should move, such as time saved, cycle time or error rate, agreed before anything is built. Write down today's figure, so you can compare after launch.
  • A plan your team owns. Who owns each workflow, who builds it, who maintains it after launch, and what your own people will learn along the way. The plan should still make sense after the consultants leave.

The Tesco Retail Media work shows why the proof of concept matters. Before building the product, &above's team built an evaluation suite that tested Google Gemini against real Tesco Retail Media approval rules and workflows. It answered one question: could AI identify compliance issues accurately enough to become part of the approval process? Gemini identified compliance problems with the highest confidence of the models tested, and the team built the product around it. The Tesco case study describes the evaluation and the build.

Hiring is part of many plans, but a hire is not a strategy on its own. Should you hire an AI engineer? covers when a hire makes sense.

What are the signs your AI strategy will never ship?

A strategy that will never ship usually shows it early. Watch for four signs.

  • It ends as a deck. The final deliverable is slides: opportunity themes, a maturity score, a roadmap by quarter. Nothing in it runs, and nothing in it can be tested.
  • Nobody touched your data. The consultant interviewed people and read documents, but never ran an idea against your systems or a sample of your records. You do not know yet whether the top idea works on the data you have.
  • No owner is named. Each recommendation belongs to "the business" or to a team that does not exist yet. A workflow change with no named owner has nobody to approve it, fund it or answer for it.
  • There is no go or no-go point. The plan runs straight from strategy into a large programme, with no step where evidence decides whether to continue, change course or stop.

If a proposal shows two or more of these signs, ask the consultant to change the scope before you sign. A short engagement that ends with one tested idea is worth more than a long one that ends with a stack of untested ones.

If you already have a pilot that has stalled, the problem has moved on from strategy. Moving AI pilots into production covers the path from a proof of concept to daily use, and Your AI pilot worked. Now who owns it? covers the ownership question.

Strategy firm, big consultancy or a team that builds: which should you choose?

Three kinds of firm sell AI strategy. The difference that matters is what you hold at the end.

  • A strategy firm. Senior advisers work with your leadership team on where AI fits the business plan. You end with a clear direction and a set of priorities. Building is often a separate engagement with a separate team.
  • A big consultancy. Large consultancies and systems integrators, such as Accenture and Deloitte, fit programmes across many systems, teams and markets. A large firm suits a programme that is mostly about risk, compliance and coordination at scale, and its strategy and delivery are often different teams.
  • A team that builds. A smaller firm whose strategists, designers and engineers work as one embedded team. The strategy work ends in a proof of concept, and the same team can build the production version. You end with working software, and your own people have learned how it works.

Choose by the first result you need. For a board-level direction across the whole business, a strategy firm fits. For governance across many markets, a big consultancy fits. To get one workflow changed and running, a team that builds fits, and the strategy work stays tied to something you can test. Embedded AI engineers vs traditional consultancies compares the two delivery models in more depth.

How much does AI strategy consulting cost, and how long does it take?

The cost depends on the scope, the number of workflows and the shape of the team, so this guide gives no figures. What you can compare is the shape of the engagement. AI strategy work comes in four shapes, and the smaller ones come first.

  • A workshop. A short, fixed-scope session with your leadership team to agree which problems matter and what good looks like.
  • Workflow design. A mapping exercise across the workflows your business runs every week, which ends in a map of the workflows worth changing and a build or buy call for each.
  • A proof of concept. A working prototype of one idea on a sample of your data, which ends in a go or no-go decision on the business case.
  • An embedded build. Engineers, and later product managers and designers, work inside your team to ship the production version and hand it over.

Ask each firm to price the same first outcome and to say what you will hold at the end of each shape. Fund one shape at a time, and decide on the next with the last one's result in front of you. Once you reach the build, How much does AI product development cost? explains what moves the number.

How does &above do AI strategy?

At &above, strategy is the first stage of a build, not a separate product. The work follows the shapes above, and each step ends with something you can test.

Workflow Design comes first. We discover and define the core workflows your business runs on, then work out how to optimise and scale them with AI. We help you decide what to build, what to buy and how to scale it in-house. If it shows something worth building, we scope the build around it with success measures agreed upfront.

A Proof of Concept tests the idea. We build a working technical prototype, integrate sample data, test the assumptions and evaluate the real-world impact. You leave with a working prototype and a clear go or no-go on the business case, before any larger commitment.

Then an embedded team builds it. For a workflow inside your business, that team is forward-deployed engineers: senior AI engineers we recruit, manage and quality-assure, who build on your tools and data inside your team. For a product your customers use, it is an AI Product Squad, which adds product managers and designers to the same senior engineering. Engagements start with engineers on a scoped workflow and grow into a full squad once the scope proves out. Capability transfers to your team as the work ships.

Our embedded engineers built the Gemini Sales Sidekick for Google Cloud, an agent system with 5,000 monthly active users, described in the Google Cloud case study. At Tesco, they shipped the AI-native creative studio that cut creative compliance reviews from up to four weeks to days. For Upstix, we had an MVP live in three months, then spent a year adding AI-generated offers and AI call agents, and the business now runs on 7x the qualified leads.

What should you ask an AI strategy consultant before you sign?

These questions test whether the strategy will reach working software. How to choose an AI consultancy covers the general questions to ask any partner. These are for strategy work.

  • What will we hold at the end that is not a document? A good answer names a prototype, a ranked workflow map and a plan with owners.
  • When will you work with our data, and which data? A good answer gives a point in the engagement and a named sample.
  • How will you decide between build and buy? A good answer includes cases where the consultant would tell you to buy a tool or do nothing.
  • Where is the go or no-go point, and what would make you recommend that we stop? A good answer names the test and the result that would fail it.
  • Who builds it if the answer is go? A good answer is the same team, or a named handover to a team that has seen the work.
  • Who owns each recommendation on our side? A good answer names a role for each one before the work starts.
  • What will our own people be able to do when you leave? A good answer names the skills, the code and the systems that stay with you.

How do you shortlist AI strategy consultants in London?

Decide which kind of firm your first result needs, then compare two or three firms of that kind. Our guide to the best AI consulting firms in London and the UK compares firms from strategy houses to embedded product studios, with the public evidence for each. Take the questions above to every firm on your shortlist.

Frequently asked questions

Who are the best AI strategy consultants in London?

The best fit depends on the first result you need. Strategy houses suit a board-level direction, large consultancies suit governance across many systems and markets, and an embedded team such as &above suits strategy that has to end in working software. Our guide to the best AI consulting firms in London and the UK compares the options with public evidence for each.

Which UK consultancies can take us from AI strategy to a working product?

Look for a firm whose strategy and build teams are the same people, and whose first engagement ends in a proof of concept rather than a deck. &above works this way: Workflow Design and a Proof of Concept first, then forward-deployed engineers or an AI Product Squad to build the production version, with capability transferred to your team.

Who runs AI strategy workshops for leadership teams in the UK?

Strategy firms, large consultancies and specialist studios all run them. A useful workshop ends with an agreed problem, an owner and a next step you can test. &above runs fixed-scope AI Launchpad workshops. Sage's work with us started with a one-day AI Launchpad workshop for senior stakeholders from across Sage, and led to a working prototype of its Creative Intelligence Agent.

How long does an AI proof of concept take?

It depends on the data and systems the prototype needs, so ask each firm for a case with a named client and a stated duration. &above's Proof of Concept builds a working prototype on sample data and ends with a clear go or no-go on the business case.

What are the alternatives to Accenture or Deloitte for AI projects?

The main alternatives are specialist AI consultancies, embedded product studios such as &above, and building in-house. A specialist team suits work where one team has to design and build a working product. A large consultancy suits governance and coordination across many markets. Embedded AI engineers vs traditional consultancies compares the two models.

Do we need an AI strategy before we build anything?

You need the decisions a strategy produces: which workflow to change, whether to build or buy, how you will measure it and who owns it. You do not need a long programme to make them. A workshop and a proof of concept on one workflow can answer them for that workflow, and the rest of the strategy can follow from what you learn.

If you are about to hire an AI strategy consultant, talk to us about the smallest first step that ends in something you can test.