The cost of building technology has long limited how much enterprises could take on. Engineering capacity was scarce, ideas waited in queues and business cases could spend months in approval. Those constraints forced choices. They did not guarantee that the right work received funding.

AI can reduce the time and cost of writing code and creating prototypes. That is useful, but a working prototype is still some distance from a reliable service that customers want and the business can support. Faster development also makes it easier to start work before anyone has properly tested whether it is worth doing.

Leaders need to ask whether an idea deserves investment and what would make them reconsider. The cost of a poor decision includes the time people spend on it and the opportunities they cannot pursue while they are occupied. Cheaper code does not remove those costs.

The decision organisations avoid

Big organisations can find it easier to start things than to stop them. Emergn recently surveyed 700 senior leaders in the UK and US for our report, The Value Gap. Only 30% said that stopping an underperforming programme was a normal part of how their organisation operated. And 71% could not give their board a complete, real-time view of their transformation and AI programmes on demand.

On average, those leaders estimated that their organisations lose 2.4% of annual revenue to transformation and AI initiatives that fail to deliver the expected value. At a company with £1 billion in annual revenue, that would be £24 million a year.

Experimentation is a separate matter. Anyone who wants to innovate must accept that some bets won't pay off. The bigger problem is work that holds on to its budget and team long after the evidence has turned against it.

The reasons are familiar. People feel that walking away wastes what has already been spent. Leaders have staked their reputations on the work. Annual budgets can reward commitment over learning, and programmes may still be judged on meeting time and budget targets without enough attention to whether they made a useful difference. It is also difficult to challenge work that leaders cannot see clearly.

Why AI makes the problem more urgent

None of this started with AI. But when it becomes easier to launch a pilot, organisations can take on more work than they can properly evaluate or support. Each initiative may look affordable in isolation. Together, they place demands on data teams, infrastructure and the people expected to change how they work.

A product team serving business customers still needs to establish whether the problem matters enough for a customer to act. A demonstration may attract interest without changing a buying decision or earning a place in the customer's daily work. Producing more features will not resolve that uncertainty.

This is where product management matters. Teams need to test their assumptions with customers, understand the cost of adoption and use what they learn to guide investment. Faster development is valuable when it helps them learn sooner or deliver something useful. Release volume alone tells leaders very little.

Making stopping part of the process

Leaders should review the case for continued investment as evidence develops. Waiting for the annual portfolio review can mean months of spending after a problem is already clear.

Funding in stages helps. Instead of approving a whole programme on day one, release enough money to answer a specific question. That might be whether a customer problem is substantial enough to solve, whether customers will adopt the proposed solution or whether the economics justify a larger commitment. The next investment should depend on what the team learns.

Agree at the outset what evidence would support continuing, changing direction, pausing or stopping. Give experiments enough time to answer their question, and assess them against that purpose. A discovery exercise should not be rejected because it has yet to produce revenue. It should be challenged if it keeps consuming money without resolving the uncertainty it was funded to investigate.

Leaders also need an up-to-date view of the portfolio, including the cost of the work and the evidence supporting it. Spreadsheets and status reports are of limited use if they leave the investment decision disconnected from what teams have learned. That principle has shaped our work at Emergn, including Praxis: helping leaders bring product evidence into decisions about what to fund, pause or stop.

Leadership makes the decision possible

Where stopping a project is treated as a personal failure, people have a reason to keep it alive. Leaders need to recognise the team that identifies a problem early and recommends stopping on sound evidence. Doing so protects money and people for work with a stronger case.

Boards should ask what the organisation has decided to stop and why, as well as what it plans to build. I would expect a leadership team to defend continued investment with the same care it gave the original proposal. A decision that made sense six months ago still needs to make sense today.