“When a man does not know what harbour he is making for, no wind is the right wind.”

— Seneca

Every few years, institutions fall in love with a tool and mistake it for a destiny. “Digital transformation” becomes the strategy, then “cloud,” and now “AI.”

The pattern is predictable. The capability feels concrete, modern, fundable, and easy to measure. The real outcome is slower, messier, harder to own, and easier to dispute. So the capability becomes the headline and the objective goes vague.

Call it the Strategy Mirage: a means pretending to be a mission. And the question the room fails to ask is the simplest one, what are we trying to win?

The cure begins with a hard distinction, because the mirage lives in the blur between two things.

An objective is an outcome in the world: safer patients, faster service, lower fraud, stronger security. An enabler is a capability that helps produce it: AI, cloud, a data platform, automation. Enablers matter enormously. They are simply not the end.

Seal this off from the genuine article, because they wear similar language. Declaring “our strategy is AI” is like declaring “our strategy is roads” instead of “faster trade and safer logistics.” Roads can be excellent. But roads do not tell you where you are going, and a capability does not tell you what you are for. The mirage is not a wording problem. It is a governance problem, because an institution that cannot name its destination will optimise the vehicle instead.

The mechanism that sustains the mirage is mundane, which is why it is so durable. Capabilities are easier to count than outcomes.

“We deployed the system” is clean and immediate. “We reduced harm” is slow, contested, and hard to attribute. So the institution drifts, almost without deciding to, toward what can be reported and away from what must be defended.

Once the capability is the centerpiece, the whole organisation reorients around delivering it. Contracts, deployments, and architectures become the work. The beneficiary’s outcome, the thing that justified the spending, becomes a downstream hope rather than the governing aim. Effort concentrates on the machinery because the machinery is legible, and legibility is what gets rewarded in a review.

The cost is an institution that can succeed completely and fail entirely at once.

This is Adoption Theater: rollouts completed, procurement closed, dashboards glowing green, while the reality the program was meant to change stays exactly where it was. If the dashboard can turn green while outcomes stay flat, the institution is not transforming. It is reporting.

The UK’s NHS National Programme for IT shows the failure at scale. After years of delay and billions in cost, it was dismantled. The lesson is not that technology is hard. It is that when a vast capability becomes the centerpiece, the institution optimises for delivering the capability, the contracts and systems, while the patient outcomes it existed to serve stay contested and unowned. You can build a cathedral of capability and still fail the mission, if the objective was never made sovereign.

The honest objection is that sometimes the tool really is the goal, and the distinction must not become dogma.

There are two legitimate cases. The first is when the capability is the product, when the technology itself is what you sell, not a means to something else. The second is when the end state is sovereign capability, building a domestic compute cluster, say, to escape dependence on a foreign supplier, where owning the capability is itself the outcome that protects you. In both, “our strategy is the technology” is true, not evasive. The test is not whether a tool appears in the strategy. It is whether the tool is standing in for an outcome the institution is avoiding naming.

The distinction is testable in one move: ask what the institution would still owe the world if the technology were already finished and installed. If a clear answer remains, lower harm, faster service, greater resilience, the technology was always the means. If the only answer is that the institution would be modern, the tool was hiding the absence of a mission.

So the operational move is to force the sentence the mirage leaves unfinished.

Make the strategy complete the clause “we are doing this so that,” and refuse to let it stop at the tool. “So that we are AI-powered” or “so that we modernise” is not a landing. The sentence must come to rest in the world: so that service time drops, errors fall, trust rises, resilience improves. If it cannot land there, you have a capability program in search of a mission.

Then apply two disciplines that close the escape routes. Name what would count as failure even if adoption is high, because “adoption is high but trust declined” is the failure the green dashboard is built to hide, and high adoption is not a moral alibi. And require that the roadmap be readable as service to a named beneficiary, with a stated order of priorities and the lines it will not cross. If the program cannot be read as service to that beneficiary, the technology is not the strategy. It is the distraction.

The simplest version of all of it is a single question, asked before the funding is approved. If the technology vanished tomorrow, would you still want the outcome? If yes, the outcome is the strategy and the tool is its servant. If the goal makes no sense without the tool, you have crowned a proxy.

So before you fund the capability, name the outcome it serves, and name the failure that high adoption could hide.

A tool can be magnificent and still take you nowhere. Decide where you are going first, then choose what will carry you.

A capability is a vehicle, not a destination; when the tool becomes the goal, you fund the means and quietly lose the end it was meant to serve.

Decision Rule — The Landing Card

Before funding any capability, tool, or transformation, force the sentence to finish in the world.

• The landing. Complete “we are buying this so that ___” until it ends on a real-world outcome, not “so that we are modern.”
• The vanishing test. If the technology disappeared tomorrow, would you still want the outcome? If no, you have crowned a proxy.
• The failure condition. What would count as failure even if adoption is high?

If the proposal cannot land the sentence and name its failure condition, treat it as procurement, not strategy, and do not call it transformation.

Name the destination before you fund the vehicle.