Your AI Strategy Is Just a Shopping List

The most common mistake is treating AI strategy as a technology selection exercise. 'We need an AI strategy' becomes 'we need to choose a platform and run some pilots.' This skips the most important step: understanding where AI actually changes your competitive position.

An adoption plan answers: what AI tools should we use? A strategy answers a fundamentally different question: how does AI change the economics of our industry, and what do we need to do about it?

I've seen this play out with painful predictability. The CEO reads an article, attends a conference, or gets spooked by a competitor's press release. They tell the CTO to 'develop an AI strategy.' The CTO assembles a working group. The working group evaluates vendors. Six months later, there's a PowerPoint deck with a maturity model, a roadmap of pilot projects, and a budget request. The board approves it. Eighteen months later, none of the pilots have scaled, the budget is spent, and nobody can articulate what the organisation actually learned.

The problem isn't execution. The problem is that nobody asked the right question at the start.

Before you can build an AI strategy, you need to understand your starting position with uncomfortable honesty. Not the version on the architecture diagram. The version that actually works in production.

This diagnostic step is the one almost everyone skips, because it reveals things organisations don't want to hear. Your data isn't as clean as you think. Your processes are more dependent on tribal knowledge than anyone admits. The manual workarounds that your team has built are load-bearing walls, not temporary fixes.

I worked with an organisation that was convinced they were 'data-ready' for AI. The diagnostic revealed that their three most critical data pipelines relied on a single analyst's Excel macros, their CRM data hadn't been deduplicated since 2019, and their 'data lake' was actually a data swamp with no governance, no lineage, and no documentation. The AI strategy they'd drafted assumed none of this was true.

The diagnostic isn't glamorous. It doesn't make for good board presentations. But without it, your strategy is built on assumptions that will collapse the moment you try to execute.

The graveyard of AI pilots is vast. Organisations run a proof of concept, it works in controlled conditions, and then it never makes it to production. The pattern is so common it has a name: pilot purgatory.

But pilot purgatory isn't a technology problem. It's an organisational choice. Pilots fail to scale because they were designed to fail.

They were built on clean data that doesn't exist in production. They required manual intervention that doesn't scale. They solved a problem that decision-makers didn't actually prioritise. They were championed by an innovation team with no authority to change operational processes. They were approved because saying 'yes' to a pilot is easy and saying 'no' requires a reason.

The uncomfortable truth is that most organisations run pilots as a substitute for making difficult decisions. A pilot lets you look like you're doing something about AI without actually committing to the organisational changes that would make AI work. It's activity theatre — and it's expensive.

Real AI strategy starts with the business problem, not the technology. It identifies specific, measurable outcomes — not vague ambitions like 'become AI-first,' but concrete targets like 'reduce underwriting decision time from five days to four hours' or 'automate 60% of customer service interactions within quality thresholds.'

Then it asks the hard questions. What capabilities do we lack? What data do we need that we don't have? What processes need to change? What roles become redundant? Who in the organisation will resist this, and do we have the leadership backing to push through that resistance?

That last question is the one that separates strategies that work from strategies that become shelf-ware. Every meaningful AI implementation changes someone's job. The organisations that succeed are the ones where senior leadership is prepared to back those changes against internal resistance. The ones that fail are the ones where the strategy was approved but the organisational change was quietly shelved because it was 'too disruptive.'

If your AI strategy doesn't name the things you're going to stop doing, the processes you're going to change, and the resistance you expect to encounter — it isn't a strategy. It's a wish list.