Practise a discovery conversation with an explicitly fictional AI role-play. Test questions and listening habits without treating simulated answers as market evidence.

Decide whether a PDF needs text recognition, interpretation or a better original file. Use a page-level check before asking AI to read it.

Create a client FAQ from genuine questions, with clear service boundaries and source-owned answers. Keep proposed questions separate from observed demand.

Build a traceable business glossary from approved documents. Separate aliases, conflicting meanings and proposed definitions before asking owners to decide.

Prepare internal instructions for AI-assisted translation with a terminology list, preserved conditions and review by someone competent in both languages.

Choose which meeting artefact the work needs: detailed transcript, concise summary or an agreed decision record. Avoid treating them as interchangeable.

Give personal voice notes a useful structure before turning them into a brief. Separate observations, ideas, commitments and questions.

Use a decision-led reading plan for long documents. Find relevant passages, preserve qualifications and keep unanswered questions visible.

Use AI to organise customer comments into traceable themes while keeping sample limits, conflicting views and original wording visible.

Describe a spreadsheet formula in plain language, define edge cases and request an explanation before using it. Includes a small worked specification.

Decide when missing information should trigger a question, a stated assumption or a stop. Includes a practical clarification ladder for business tasks.

Frame an AI spreadsheet analysis around a decision, population and comparison. Use a question card before uploading data or accepting a chart.

Choose between finding information, preparing a draft and changing a system. Use a task-boundary worksheet before selecting an AI tool.

The new version excels on the demonstration example. Someone declares it better. Nobody checks the awkward requests that caused problems with the current version.

The builder is on leave. A source document changed, cases are failing and nobody knows whether to retry or stop. The workflow depends on one person's availability.

The AI recommends moving a deadline. The next step updates the project and sends an email. Somewhere between a suggestion and an external commitment, the decision disappeared.

Drafting becomes faster, so the team increases volume. The review queue grows faster than anyone can clear it. The apparent capacity gain disappears into waiting and rushed approvals.

The summary says somebody will contact the supplier. Nobody remembers agreeing who. The AI has created a tidy task list from a discussion without a complete commitment.

An AI email sounds polished, so the reviewer sends it. The message promises a date the team has not agreed. The tone check worked. The approval process missed the commitment.