The facts

The study published by ActivTrak tracks digital work activity across more than 9,000 employees in 467 companies. Contrary to the promise of shorter workdays, the data shows that regular AI users actually log longer hours: their workday averages between 8.4 and 9.1 hours, compared with roughly 7.6 hours for non-users — up to 90 extra minutes per day.

Among heavy users (more than 45 minutes across at least five daily sessions, often technical or operations roles), activity spills further outside standard hours. Two mechanisms drive this paradox: the “verification tax” (reviewing, fixing and fact-checking generated outputs) and the volume effect, where producing more drafts or code shifts the bottleneck to review and approval steps.

Editorial view

Generating a draft faster does not shorten a workday if proofreading and copy-pasting remain manual. Across the real-world cases documented on IA du Jour, net time savings appear when automation targets a specific repetitive task (PDF extraction, quote drafting, ticket triage) with clear human sign-off at the end of the pipeline.

What this changes in practice. Before multiplying one-off prompts, pick one measurable weekly task and structure the final review, as outlined in our guide Automate tasks with AI.