AI Strategy Consulting: Why So Much of It Never Ships

I will tell you the part that most AI strategy consulting pages leave out. A great many of these engagements end with a polished slide deck and nothing actually running in the business. The owner pays for the thinking, files it away, and six months later sits roughly where they started, only lighter in the bank. I have watched it happen often enough to find it genuinely frustrating. The strategy itself was usually fine. What went missing was the bridge from the slide to a working system.

So here is the honest version. I want to walk through what this work is meant to do, and the handful of ways it quietly gets wasted, so you end up with a strategy you can actually build and own. I run this work for a living, so I am not pretending the role has no value. The value is just easy to waste, and worth protecting.

What AI strategy consulting is supposed to do

Stripped of the jargon, AI strategy consulting answers two questions. Where could AI realistically help this business, and in what order should you go after it. A proper engagement starts by looking hard at how ready you are and what state your data is in. From there it names the use cases worth pursuing and puts them in a sensible order, then settles the practical questions of tools and governance.

That backbone is shared by everyone serious in this space, from a UK firm like help4IT through to the big consultancies. The disagreement is not about the ingredients. It is about how much of it ever turns into something that runs.

Where AI strategy consulting quietly goes wrong

Here are the failure modes I see most, and what each one actually costs you.

1. The strategy ends as a deck nobody builds. This is the big one. Most pricing stops at the roadmap, and several firms openly hand the build to a separate partner. The handover is where momentum dies. You are left holding a document that describes value rather than a system that delivers it.

2. It tries to boil the ocean. A sweeping transformation plan looks impressive and is almost impossible to start. The fix is unglamorous. Pick one high value use case, ship it, and let the proven return earn you the right to the next one. Automating invoice processing or first line customer replies pays back faster than a company wide programme. It also builds momentum you can feel.

3. Nobody checks the data first. AI is only as good as what you feed it. BCG’s own framework, which it calls the 10, 20, 70 rule, puts roughly 10 percent of the value on the algorithms, 20 percent on the technology and data, and a full 70 percent on people and process. If your data is messy and your team has not changed how it works, the cleverest model in the world will underwhelm. It is also why so many projects miss. Reporting in early 2025 found only about a quarter of companies seeing a clear return on AI.

4. You are sold a retainer, not a capability. Plenty of engagements are built so the dependency never ends. The build sits with the consultant, the documentation is thin, and the monthly fee keeps arriving. There is nothing wrong with ongoing support you choose. There is a lot wrong with support you cannot leave.

5. Governance gets bolted on at the end. Basic privacy and record keeping are cheap to design in and expensive to retrofit. For a UK business this is not optional. UK GDPR and the ICO apply the moment personal data is involved, and the EU AI Act reaches any business selling into the EU. Bake it in from the first use case.

What a good AI strategy actually delivers

Flip every failure mode above and you get a useful checklist. A good engagement hands you a short, prioritised list of use cases rather than a wish list, and a roadmap tied to your real goals with the first item small enough to start on Monday. It names the tools plainly, the everyday ones the big firms themselves lean on, like ChatGPT Enterprise and Microsoft Copilot, alongside the automation platforms and Claude Code that let you build the connective logic yourself.

Your data gets treated as a first class problem, not a footnote, and governance stays simple enough to actually follow. Then comes the part most pages skip. At least one automation is left running before anyone signs off. A strategy you can see working is worth ten you can only read.

The other quiet marker of a good engagement is what happens to the knowledge. If everything the consultant learned about your business walks out the door when they do, you bought a dependency. If your team comes out understanding the systems and able to extend them, you bought a capability. That distinction is the whole game. It is the reason I work the way I do, and you can read more about why I teach rather than lock in.

What it costs, and where the money goes

Pricing is where the tiers separate hard. At the top, an MBB or Big Four programme is built for enterprises and priced for them, easily into six and seven figures across a multi year engagement. At the other end, an independent or boutique can deliver a real strategy and a first working use case for a small business at a fraction of that.

As an indicative international guide for 2026, a fixed fee readiness review tends to land around £8,000 to £20,000, a strategy with a roadmap around £20,000 to £60,000, and a pilot build higher again. Ongoing managed support, where it exists, often runs from a few thousand pounds a month upward. Those figures come from secondary 2026 data that is mostly US derived, so treat them as a shape rather than a quote, and remember UK and European rates usually sit a little below the US equivalents. The honest takeaway for an owner of a five to fifty person business is simple. You are almost never the customer the enterprise pricing was written for, so do not pay as if you were. If you want a sense of your own numbers first, my free AI Opportunity Scorecard gives a quick estimate of the hours and money slipping away each month.

Could you do the first pass yourself?

Here is the question almost no ranking page wants to ask out loud. Do you even need to outsource the strategy. For a small business, a lot of the early work is within reach. You can list where your team repeats itself, pick the single most painful task, and build a first version with tools that did not need a developer a year ago. Claude Code in particular has dropped the barrier far enough that a non technical owner can ship a real workflow. That is exactly the ground I cover in my hands on work.

That does not make consultants pointless. An outside view is genuinely useful for spotting the highest value opportunity and for the messier builds that span several systems. The shift I argue for is one of ownership. Bring someone in to accelerate and to teach, not to hold the keys. Automation is one half of how I help businesses grow. The other is search and AI visibility, making sure you are the name that gets recommended across Google and AI tools when buyers go looking.

So my honest advice on AI strategy consulting comes down to one test. At the end of it, do you have a system running and the ability to run it without the person who built it. If the answer is a deck and a retainer, keep looking. If the answer is working software and a team that understands it, you have bought the right thing.

If that is the outcome you want, start by seeing your own numbers, then build something real and small. Take the AI Opportunity Scorecard for a two minute estimate, or book an AI Opportunity Audit and we will map your best first use case together, then build it and leave you able to carry it on yourself.