Begin with a workflow people recognise
An announcement about “becoming an AI-driven organisation” leaves many practical questions unanswered. Start with a specific workflow: preparing meeting notes, finding internal guidance or producing a first draft of routine material. Explain why that task is being considered and what a useful improvement would look like.
Invite the people who perform the work to describe its difficult parts. A task that appears repetitive from a distance may include exceptions, relationships and judgment that are easy to miss. Understanding these details before choosing a tool makes the trial more relevant to the team.
Make the pilot explicit and limited
A pilot should have a defined scope, an owner and a review date. Clarify what information may be used, how outputs will be checked and which actions still require approval. Do not promise that the trial will have no effect on roles if that question has not been decided; explain what is known and how changes will be discussed.
- Choose one task and a small group of willing participants.
- Record the current effort, quality issues and review steps.
- Agree which data and tools are approved for the trial.
- Provide time for practice and a place to report problems.
- Include checking and correcting outputs when measuring the workload.
For example, drafting meeting summaries may reduce writing time while increasing the need to check decisions and ownership. Measure the whole process, including corrections, rather than celebrating the speed of the first draft alone.
Treat feedback as part of adoption
People need a way to raise concerns without being labelled resistant to change. Ask participants what helped, what created extra work and where the output could not be trusted. Distinguish a training gap from a task that is unsuitable for the tool.
At the review, choose whether to continue, adjust or stop the pilot. Share the reasons and the evidence, including disappointing findings. If the trial expands, update responsibilities and training rather than assuming the original arrangements will scale automatically.
Trust is supported by consistency between the announcement and the experience. Keep the promises you make about consultation, review and responsibility. A smaller rollout that people understand is a more useful starting point than a broad requirement nobody knows how to apply.
