The idea

Use AI to do the paid work you already do faster or better.

Who the evidence applies to: People in office and knowledge jobs whose work involves writing, customer support, analysis or routine documents, with the largest measured gains for less experienced workers.

Employers and the UK government are rolling out AI assistants to staff, and studies of their real effect on productivity continue to be published.

How we checked

This is an evidence check: we looked for large studies, official statistics and published datasets, opened every source, and had a second researcher check each figure against them. 4 sources, listed below.

Costs

Start-up cost
Varies
Effort
low

What the evidence shows

Controlled studies consistently find that generative AI makes many office tasks faster and sometimes better, with the biggest gains for less experienced or lower-performing workers. A study of 5,179 customer support agents (Brynjolfsson, Li and Raymond) found a 14% average rise in issues resolved per hour, and 34% for novice and low-skilled agents, with minimal impact on experienced staff. In a Harvard/BCG experiment with 758 consultants, those using AI on suitable tasks completed 12.2% more tasks, 25.1% more quickly, at over 40% higher quality, but on a task outside the AI's abilities they were 19 percentage points less likely to get the right answer. None of these studies shows the time saved turning into higher pay.

What to expect: You can reasonably expect to save time on writing, drafting and routine tasks, but the gains are uneven, depend on the task, and need checking because AI can make confident mistakes. Any pay benefit depends on your employer or clients rewarding the extra output, which the evidence does not show happens automatically.

First steps

  1. Pick one routine task you do often, such as drafting emails or summarising documents, and try doing it with an AI tool your employer allows
  2. Check every output yourself, especially for tasks that need judgement or specialist knowledge, where studies found AI can lead people astray
  3. Keep a simple record of time saved and quality so you can show the difference
  4. Use that record to make the case for taking on more or better-paid work

Limitations

Studies report averages: your own result depends on your field, skills and where you live. Figures come from the sources below and may be updated as new evidence appears.

VerdictWorks with limits

Works for some, not most.

Sources

  1. Brynjolfsson, Li and Raymond, 'Generative AI at Work' (NBER working paper; Quarterly Journal of Economics, 2025): 5,179 customer support agents; 14% more issues resolved per hour on average, 34% for novice and low-skilled workers, minimal impact on experienced workers nber.org
  2. Noy and Zhang, MIT experiment with 444 college-educated professionals doing writing tasks: ChatGPT cut time taken by 10 minutes (37%) and raised output quality by 0.4 standard deviations, helping lower-ability workers most economics.mit.edu
  3. Dell'Acqua et al., Harvard/BCG 'jagged frontier' experiment with 758 consultants: 12.2% more tasks, 25.1% faster and over 40% higher quality on tasks within AI's capability; 19 percentage points less likely to be correct on a task outside it; lower performers gained 43% versus 17% for higher performers mitsloan.mit.edu
  4. UK government Microsoft 365 Copilot experiment with 20,000 civil servants across 12 organisations (Sept–Dec 2024): participants self-reported saving an average of 26 minutes a day; the tool struggled with nuanced work and needed human oversight assets.publishing.service.gov.uk