Language and display
BE THE BOSS OF AI.ADRIAN DUNKLEY ↗
← This newspaper edition

Where could AI help your team? Start with one task

A US poll found a gap between jobs that can be done remotely and jobs that require people on site. A workplace can use the distinction to choose a fair trial.

A black pen lies across the blank page of an open spiral notebook on a dark wooden desk.
Desk photograph by Gabriel Cox / Unsplash · Licence

In a Gallup report published on 16 November, 49% of US employees whose roles could be done remotely said they used AI at work at least a few times a week. Among employees who had to work on site, 22% reported using it that often. Gallup surveyed 2,648 US employees from 5 to 19 August. A job was classed as remote-capable if staff could do it remotely, even if they worked in an office. The survey did not measure the quality of their work or the frequency of AI use in the Caribbean.

Start where the team can compare work fairly

At a tour operator, preparing a written booking confirmation is different from meeting a guest at a pick-up point. For a small trial, use the confirmation. Give two colleagues these practice bookings. Case A: 25 November at 09:00, East Gate, full refund for cancellations by 09:00 the previous day, practice contact 202-555-0101. Case B: 26 November at 11:00, West Gate, full refund by 11:00 the previous day, practice contact 202-555-0102. For each case, one colleague uses the usual template and the other uses the same template with an approved AI assistant. They work independently and swap methods for Case B. Record any wrong or missing details in the first draft. Measure time from opening the booking to approval, including corrections. Keep the current template if the assisted draft is faster to write but the overall work still takes longer or the confirmed details are wrong. Two practice bookings can reveal a defect. They cannot establish savings across the team.

Put the task list ahead of the job title

The AI Job Exposure Score asks staff about routine writing, computer-based work, in-person judgement and employer adoption. Use a recent week rather than an ideal week when you select answers. It reports a score and a breakdown by area. A practice run choosing the second option on all ten questions showed 33 out of 100 and “Augmentation Zone”, its label for a role where a tool may assist with some tasks. This self-assessment cannot forecast job losses. Use the routine-work breakdown to choose one task to evaluate, then use the time and error records to decide whether it belongs in the next staff session.

Sources and further reading

Sources checked 18 November 2026.