The gap between AI strategy and implementation is where billions of pounds go to die. Here's why, and what to do instead.
In 2024, global spending on AI consulting exceeded $20 billion. By any reasonable measure, the return on that investment has been catastrophic. McKinsey's own research suggests that fewer than 20% of AI initiatives move beyond pilot stage. Boston Consulting Group puts the failure rate of large-scale digital transformations at 70%. These are not fringe statistics from pessimistic analysts. These are the consultancies' own numbers about their own industry.
The pattern is always the same. A company hires a Big Four or MBB firm. A team of analysts arrives, conducts interviews, reviews data, and produces a strategy document. The document is beautifully formatted. It contains a maturity assessment, a roadmap, a capability framework, and a set of recommendations. It costs somewhere between 150K and 2M. And then it sits on a shelf.
Why? Because the people who wrote it cannot build the systems they recommended. They are strategists, not engineers. They can tell you that you need an AI pricing engine, but they cannot build one. They can recommend a customer data platform, but they have never deployed one. The strategy is correct in the abstract and useless in practice.
Let's be specific about what you get for a six-figure strategy engagement. You get a current-state assessment that tells you things you already know, expressed in consultant language. "The organisation exhibits a low level of AI maturity across key business functions." Translation: you are not using AI yet. You paid 200K to hear that.
You get a target-state vision that describes a future so generic it could apply to any company in any industry. "Leverage AI-driven insights to optimise customer lifetime value through personalised experiences." This is not a plan. This is a sentence from a LinkedIn post.
You get a roadmap with phases. Phase 1: Foundation. Phase 2: Scale. Phase 3: Optimise. Each phase has a timeline measured in quarters or years. None of them have engineering specifications. None of them have cost estimates for the actual build. None of them name the specific systems, databases, APIs, or infrastructure required.
And you get a capability framework with a maturity model. You are currently at Level 2. You want to be at Level 4. The steps between those levels are described in terms so vague that no engineer could act on them without starting from scratch.
The strategy-only model persists because it is extraordinarily profitable for consultancies. A team of three analysts can produce a strategy document in 8-12 weeks. The firm bills at 2,000-5,000 per day per person. The total cost is high, the delivery is a PDF, and there is no accountability for implementation.
If the recommendations fail when someone else tries to implement them, that is an implementation problem, not a strategy problem. The consultancy's work is done. They have moved on to the next client.
This creates a perverse incentive structure. The consultancy benefits from complexity, not simplicity. A simple recommendation — "replace your pricing vendor with an in-house engine, here's the code" — would be too easy to evaluate. A complex, multi-year roadmap with dozens of interdependencies is impossible to evaluate until years later, by which time the consultants are long gone.
The client, meanwhile, is left holding a document that requires a second round of hiring — either internal engineers or a systems integrator — to actually do anything with. The total cost doubles or triples. The timeline extends by 12-18 months. And by the time anyone builds anything, the original strategy is outdated.
Curious what your margin opportunity looks like?
Free Tool
How much margin are you leaving on the table?
Answer 6 questions. Get a personalised margin estimate in under 2 minutes.
Take the Free Margin AuditWe know what works because we have done it. At a major UK fashion e-commerce retailer, we designed and are executing a £6.4M transformation programme with 119 initiatives across 7 workstreams. Every initiative has an owner, a go-live date, an EBITDA value, and a status indicator.
The difference is not just in the planning. We built the systems ourselves. We wrote the AI pricing engine that delivered +77% revenue uplift. We built the customer data platform that processes 16.4 million profiles. We deployed seven AI agents that run across every department. We migrated the cloud infrastructure and cut hosting costs by 50%.
No strategy document preceded any of this. We audited the business, identified the opportunities, built the systems, deployed them, measured the results, and reported to the board. The entire cycle from audit to live system was weeks, not quarters.
This is what implementation looks like. It is messy. It involves debugging database queries at midnight, negotiating with vendor APIs that do not work as documented, and rewriting pricing logic when the first model does not generalise across product categories. None of this appears in a strategy deck.
Before hiring any AI consultancy, ask three questions.
First: Can you show me a system you built that is running in production today? Not a proof of concept. Not a demo. A production system that is processing real data and making real decisions for a real business. If the answer is no, you are hiring a strategy firm, not an implementation partner.
Second: What is your deployment timeline for the first initiative? If the answer is measured in quarters, walk away. A competent team can deploy a meaningful AI system — a pricing engine, a customer segmentation model, an automated monitoring agent — in weeks. Our pricing engine went from concept to live deployment in under six weeks. The CDP was operational within a similar timeframe.
Third: How do you measure success? If the answer involves maturity models, capability assessments, or readiness scores, you are being sold a process, not an outcome. The only metrics that matter are financial: revenue generated, costs reduced, margin improved. We track every initiative against EBITDA. The board sees real numbers, not maturity scores.
The consultancy model is under pressure from two directions. From below, AI tools are commoditising the analytical work that justified high day rates. A competent operator with Claude, Cursor, and access to company data can produce a current-state assessment in days, not weeks. The information asymmetry that consultancies relied on — "we know things you don't" — is evaporating.
From above, boards and PE firms are demanding measurable ROI from AI investments. "We completed a strategy engagement" is no longer an acceptable update. "We deployed a pricing engine that generated 734K in annual value" is. The shift from activity-based to outcome-based measurement is killing the slide-deck model.
The consultancies that will survive this shift are the ones that can build. Not advise. Not recommend. Build. Write the code. Deploy the infrastructure. Train the models. Monitor the systems. Own the outcome.
I built MarginOps on this principle. I've built, scaled, and sold technology companies. I don't produce strategy documents. I produce production systems with EBITDA values attached to every one. If you're tired of paying for slide decks, let's talk.
We'll identify the highest-impact opportunities in your business and show you exactly what we'd build, what it would cost, and what it would return.
We go into businesses and make them permanently more profitable. Every initiative is EBITDA-tracked.