Strategy & Roadmap

The business that doesn't know what to do with AI rarely lacks ideas. It usually has too many: a vendor demo someone liked, a feature a department head keeps asking for, a list of use cases from an offsite. What it lacks is a decision about which of these are worth doing, which aren't, and in what order. The providers who work on this problem define it that way. Wirral.ai's starting diagnosis is AI projects that begin with a tool and then go looking for a problem. Charlene Li sees organisations holding lists of AI use cases that never turn into a strategy. The remedy they describe isn't a longer list. It's a narrower commitment.

A good strategy is mostly refusals

The most useful change in thinking these providers offer is that the value of strategy work lies in what it rules out. Wirral.ai argues that the most valuable deliverable can be a list of what not to automate. Deep Purple designs its readiness assessment to hunt for reasons not to build, not only reasons to proceed. Appfront's discovery work ends with a verdict on where the client isn't ready. IT Desk will tell a client that the feature they want isn't the highest-value one to build.

This matters to a buyer because it runs against the commercial logic of most advice. An adviser who is paid more when there is more to do has little reason to say "don't." LogicRoad describes its practice as recommending the tool you already have, even when that means less work for LogicRoad. GrowExx applies the same discipline before the engagement starts: it will say no before signing if 90-day ROI isn't achievable. When you evaluate a strategy partner, one clear signal is whether they can show you the things they talked a client out of.

What readiness actually means

Refusal only has value if it rests on a clear view of what makes an opportunity ready. The providers here move readiness away from an organisation-wide score and down to the level of the individual process.

Chris Talintyre treats a single workflow as the unit of AI economics, rather than a department. He scores each workflow on how specific it is and how mature the market for solving it is, and uses that to decide which lever to pull. AI Solutions for Small Biz applies a similar test from a different angle: it scores each workflow by data readiness before anyone commits to automating it.

Kuldeep Deshpande adds a point about sequence. Many firms first take an inventory of their data and only then ask what to do with it. He argues for running data inventory and use-case discovery in parallel, so that each informs the other and neither becomes a bottleneck.

Other providers push readiness past data and into how the work is actually done. Ronins holds that you should fix the broken process before applying AI to it, because automation speeds up whatever is already there. Wirral.ai sets an equally blunt test: if nobody can name who checks the output, the use case isn't ready to roll out. Evolution Analytics widens the point, arguing that AI success depends more on how it's used than on how it's built. Moniek James offers a concrete symptom of the same issue. Generic AI output, in her view, points to a voice the organisation has never defined, not a gap in its tools. That is why her practice defers tool selection until decision architecture is in place.

Together, these views mean readiness isn't something a buyer has or lacks across the board. It holds for some workflows and not others, and a good assessment tells you which are which.

Should the people who plan also build?

This is where the providers genuinely disagree, and the choice matters for any buyer.

One group argues that strategy loses its connection to reality when it is separated from delivery. OpenKit insists that the team that writes the roadmap builds the first workflow. UnifyAI keeps the same advisor from plan through go-live, in response to AI reports that end up in a drawer. Winder.AI has strategy and production code written by the same people. Bell Integration explains why this continuity matters: it treats deployment as riskier than the proof of concept, not safer. If that is true, a roadmap written by people who never face deployment will consistently underestimate where the difficulty lies.

The opposing concern comes from Acuity AI Advisory. Its starting point is AI advice shaped by a vendor's product catalogue instead of the client's situation. Its answer is independence built into the business model, not just claimed in the pitch.

Both concerns are legitimate. Continuity protects against roadmaps that can't be built. Independence protects against roadmaps that happen to require what the adviser sells. Which risk is larger depends on how much the buyer can judge for themselves. If you choose a provider who builds, the refusal behaviours described above become your main protection. If you choose an independent adviser, check that someone will carry the plan through deployment.

Guarantees, and what they measure

Several providers tie their fee to results. Cork AI Consulting charges no fee unless five hours a week of savings are found. heybrb.ai makes the same commitment in its readiness assessment for small businesses. Automaly refunds the fee if fewer than five positive-ROI opportunities turn up.

For a small business unsure whether AI is worth exploring at all, these guarantees lower the risk of finding out. A careful buyer should still notice what they reward: finding opportunities. That is useful, but it pulls in the opposite direction from the refusal logic above. The strongest outcome of an assessment might be fewer, better-grounded commitments. When you take up a guarantee, ask the provider how they would report a workflow that clears the savings threshold but fails the readiness tests.

A plan that outlives the engagement

The last question is what happens once the strategy document exists. Charlene Li argues for an 18-month rolling plan, re-baselined quarterly instead of annually. On this view, the roadmap is a working instrument that gets revised as evidence comes in, not a fixed commitment.

Put together, these practices describe what good strategy work looks like. It tells you what not to do. It judges readiness workflow by workflow and includes the human and process conditions, not just the data. It is honest about who will carry the plan into production. And it leaves the organisation able to revise the plan without the adviser in the room.