Resources

Your AI pilot has stalled: where is the blockage?

A pilot can produce a convincing demonstration and still be difficult to use. Before adding a tool or changing models, examine the problem, access to information, fit with the workflow and the decision expected.

Updated: 12 October 2026

The problem changes at every meeting

If everyone expects a different output, the scope may be too open. Walk through one example with its user from start to finish. Write down what the pilot prepares, what the human decides and what is out of scope. Agree this before another iteration.

Data preparation takes too much work

Conflicting versions, missing documents or unstable access rights can consume the planned effort. Measure this preparation separately. Identify the authoritative source and its owner. If essential information does not exist, the project must organise its collection first.

The demonstration skips difficult steps

Observe a user without preparing the path for them. Must they copy data, reformat output or find sources themselves? Check exceptions and interruptions. One feature can work well while the whole process takes more effort. Reduce manual hand-offs before expanding the pilot.

Nobody can decide what comes next

Clarify who accepts the result, funds operation and handles incidents. Rebuild the decision criteria using available observations. Propose a limited next test or documented closure. Stopping an insufficiently useful opportunity can release resources for a better-defined need.

Your checklist

Further reading

One process. One next step.

Start with your daily work to decide what is worth checking together.