Every time someone asks ChatGPT to tidy up an email or Gemini to summarise a report, a small amount of electricity gets used somewhere in a data centre. On its own, that's negligible but multiplied across an entire workforce, it starts to add up to something worth measuring.

New research from Uswitch Business Energy estimates that across the UK, teams complete around 22 million AI prompts every working day, drawing roughly 1.37 gigawatt-hours (GWh) of electricity a year in the process - enough to power 508 UK homes for a year.

The research, based on a survey of workers combined with YouGov and workforce data, found that the average team now completes around 24 AI-assisted tasks every working day: drafting emails, summarising documents, analysing spreadsheets, automating admin. IT and telecoms teams lead the way, averaging 35 tasks a day, followed by finance with 29 a day. 

As expected, there’s also a generational pattern. Workers aged 25 to 34 report the heaviest AI use, averaging 30 tasks a day between their teams, compared with just 18 daily tasks among those 55 and over, with AI adoption in the workplace following the same age curve as most technology shifts before it.

While a single AI request uses only a small amount of electricity, the cumulative impact is significant. The scale of adoption means businesses are collectively generating billions of AI interactions every week. "As more organisations adopt AI to support tasks such as content creation, customer service and administration, the cumulative energy demand from these tools will continue to grow," says Ben Gallizzi, energy expert at Uswitch Business Energy. "Businesses already carefully monitor energy use from equipment, heating and lighting. As AI becomes more deeply embedded in workplace processes, organisations may also need to consider how digital tools contribute to their overall energy footprint."

Australian businesses have been just as quick to fold generative AI into daily workflows, from marketing teams to customer service desks, potentially without visibility into what that shift costs in electricity. As AI tools multiply across more platforms and more departments, the gap between "invisible productivity boost" and "measurable business expense" is likely to narrow.

Gallizzi's advice for businesses wanting to get ahead of it is fairly simple: audit which AI tools are actually in use across teams, consolidate where there's overlap rather than running several platforms at once, and start factoring AI use into energy forecasting the same way equipment already is. Training staff to write use AI efficiently and write prompts that produce useful outputs first time rather than iterating through multiple attempts, can also reduce unnecessary processing.

It's a small shift in thinking, but an inevitable one. AI has moved from novelty to infrastructure in most workplaces in under two years, and the energy bill that comes with it is worth watching.