How to Actually Land an AI Strategy
An AI strategy rarely fails on the technology. It fails on the organization around it — no clear mandate, no link to company goals, no way to show what the work was worth. Here is a five-step approach that puts those things first.
Most conversations about AI strategy start with models and tools. In my experience that is the wrong end. The initiatives that stall rarely stall because the technology couldn’t do the job. They stall because nobody had a clear mandate, because the work was never tied to a goal the company actually cares about, or because — six months in — no one could show what any of it was worth.
So the approach I use puts the organization first and the technology second. It has five steps, and although I’ll list them in order, in practice they interleave: you learn something about a use case that changes your view of the foundation, and you loop back.
1. Get your mandate
Before anything else, get an explicit mandate — and make sure it is the right mandate. Specifically, make sure it includes looking at organizational approaches, including how people collaborate, and not just at buying technology. An AI initiative that is only allowed to touch the tech stack, and not how the company is set up to use it, is missing most of the levers that decide success.
2. Identify the goals
Tie every AI initiative to a company goal. This is the single most useful discipline in the whole method, because it is what keeps the work visible and funded. If each project or product can point to the company goal it serves, the relevance of your AI portfolio stays obvious to the people holding the budget.
It helps to know the small set of goals AI initiatives usually serve. I work with six:
- Efficiency — the same result with fewer resources, or a better result with the same resources.
- Turnover — enabling or growing revenue.
- Quality — better outcomes, fewer errors.
- Agility — responding to change faster.
- Innovation — making genuinely new things possible.
- Employer branding — being a place skilled people want to work.
Naming the goal a use case serves sharpens it immediately, and it makes the later job of measuring impact far easier.
3. Build the organizational and technical foundation
This is the step that gets skipped, and it is the one that matters most.
On the organizational side:
- Build a cross-functional core team — data and AI specialists, developers, domain representatives, and Enterprise Architects who understand how the company’s parts, its data, and its ideas actually connect. Add a strong sponsor.
- Establish a decision-making body so the initiative has real authority and a clear path to yes.
- Embed AI into the company’s existing project and product processes: how budget gets approved, how work is prioritised, how it is governed. AI that lives in a side process stays a side project.
- Make progress visible, and invest in internal communication so staff and management understand what is happening and why.
On the technical side, the recurring theme is that data has to be able to flow through the company and be available in time. That means deciding how data is provided (centrally, decentrally, or a mix), making existing data structures discoverable, keeping data quality high, and giving teams both AI use cases and the technical foundations to build their own.
4. Identify and realise use cases
Now collect use cases — and be deliberate about sequencing them. Identify a few quick wins and lighthouse projects early. Their job is not only to deliver value but to keep management support high while the harder, slower work matures. One caveat I’d add from experience: aim your lighthouse projects at real, customer-facing value, not just at internal templates. A lighthouse that only impresses the people who built it doesn’t hold attention for long.
5. Measure the impact and make it transparent
Finally, make the value visible — including the honest, awkward parts.
A simple technique works well here. For each initiative, tag how much AI actually contributed:
- Would not be possible without AI.
- Would be noticeably less successful without AI.
- AI’s contribution is irrelevant.
This is more useful than a single headline number, because it tells the truth in both directions. It shows where AI is genuinely decisive — and it stays credible by admitting where most of the value came from somewhere else. That honesty is what lets the numbers survive scrutiny, and surviving scrutiny is what keeps the next year funded.
The through-line
Read back over the five steps and you’ll notice the technology is only one of them. Mandate, goals, organization, sequencing, and honest measurement do most of the work. That is not an argument against taking the technology seriously — it is an argument for organizing data and AI around the value the business actually creates, rather than treating it as a clever silo off to the side. Get that right, and the models have something worth doing.