The short version: most people at work in 2026 are not using the AI tool their company paid for, they are using their own ChatGPT or Gemini account on their phone because it is faster and nobody is watching. The real daily use is boring: first drafts, meeting notes, summarising long documents, and rewriting things to sound less stiff. The interesting story is not the technology, it is how much of it is happening off the books.
The gap between what's approved and what's used
Here is the bit that gets left out of most of these round-ups. Companies spent 2024 and 2025 rolling out Microsoft Copilot licences, Google Gemini for Workspace, bespoke internal chatbots built on top of Azure or AWS, all sitting inside proper procurement, security review, the lot. And a huge chunk of that spend is sitting unused, or barely used, while the same employees are pasting client emails and internal reports into their own personal ChatGPT account on their phone during their lunch break.
Microsoft's own Work Trend Index found the majority of employees who use AI at work bring their own tools rather than wait for IT to approve one. That was true two years ago and it is still true now. I see it constantly with clients: a finance director proudly shows me the new AI assistant they've licensed for the whole team, then two desks over someone admits they still use their personal account because the company one is locked down, slow, or forgets everything between sessions. Shadow AI use is the real story of 2026, and almost nobody puts it in a headline because it makes the big rollout budgets look a bit silly.
What a normal Tuesday looks like
Strip away the case studies about "AI transforming the enterprise" and here is what people do, hour by hour:
- First thing, someone pastes yesterday's meeting transcript from Otter.ai or Fireflies into ChatGPT and asks for five action points, because nobody reads the full transcript
- A sales manager drafts a follow-up email, gets Claude or Copilot to "make this shorter and less needy," and sends it with barely a tweak
- A marketing exec asks AI to summarise a 40 page competitor report into six bullet points before a 2pm meeting
- An HR manager drafts a difficult conversation script with AI before a redundancy meeting, then deletes the chat afterwards
- Someone in customer service pastes an angry customer email into AI and asks for a calmer reply, tweaks the tone, sends it as their own
None of that is glamorous. None of it is "AI transforming the workplace" in the way the vendor decks promise. It is small, repeated, unglamorous time-saving, done quietly, often without telling a manager how the draft got written.
A client story: the agency that cut its proposal time by 70 percent, and lied about it for months
I worked with a training company in Leeds last year, twelve staff, decent turnover, drowning in proposal writing. Their business development manager was spending roughly six hours per proposal, pulling old documents together, rewriting case studies, adjusting pricing tables. We built her a simple prompt library in ChatGPT, nothing clever, just a set of saved instructions for tone, structure, and the company's usual objections.
Within three weeks her proposal time dropped from six hours to under two. That is not a guess, she tracked it herself in a spreadsheet because she didn't quite believe it either. Here is the uncomfortable part: for the first two months she did not tell her boss how she was doing it that fast. She was worried it would look like she'd been slow before, or that it would make her job look replaceable. That fear is more common than anyone admits in public. People are not just quietly using AI at work, plenty are quietly hiding how much they use it, because they are worried about what it says about the job they were doing before, or what it might mean for the job they're doing next.
Where AI is replacing paid hours, not just speeding them up
This is the part that matters most for anyone earning money from their skills rather than a salary. In three areas I watch closely, AI has moved from "helpful tool" to "replacing billable hours":
Freelance writing is the clearest one. Clients who used to pay £150 to £300 for a 1,000 word blog post now expect a writer to produce a first draft with AI and charge for editing, structure, and judgement instead. If you're building income from freelance writing work, the money has shifted from typing speed to editorial judgement, and clients notice fast if you're just pasting out raw AI text with no shaping.
Virtual assistant work has changed in a similar way. The admin tasks that used to fill a VA's day, drafting emails, summarising calls, formatting reports, are now half-done by AI before the VA even opens the document. The VAs earning well in 2026 are the ones who use AI to do the boring 70 percent faster and spend the time they've freed up on judgement calls a client needs a human for. If you're starting out, the practical steps in this guide to working from home as a virtual assistant cover exactly where that shift is happening, and the guide to getting hired as a VA in 2026 is blunt about which tasks clients now expect AI to have already handled before you touch them.
Customer service is the third one, and it's the most uneven. AI drafts the reply, a human still has to send it, own it, and deal with the fallout if it's wrong. Some companies have quietly cut headcount here, others have found AI drafts create more work because agents now spend time correcting a confidently wrong AI answer rather than writing their own. The honest picture of pay and burnout in this space, including where AI has helped and where it's made things worse, is laid out in this look at customer service work from home jobs.
Underneath all of this sits a bigger shift in how people even find these jobs and gigs in the first place. Search behaviour itself has changed, more people ask ChatGPT or Perplexity directly rather than typing into Google, which changes how work gets advertised and found. I wrote about that shift in detail in this piece on AI search demand in 2026, and it's worth reading if you're wondering why some job boards and gig sites feel quieter than they used to.
A five-step way to build a habit, not just poke at it
Most people I coach through this don't fail because the tool is bad. They fail because they never build a repeatable habit, they just fiddle with it once a fortnight and forget. Here is what works, in order:
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
- Pick one recurring task first, not five. Meeting notes, or client emails, or blog first drafts. One thing you do every week already.
- Write the prompt down and save it. Not "summarise this," but the full instruction: tone, length, what to leave out, what format you want it in. Save it somewhere you'll find it again.
- Run it for two weeks and time yourself. Like my Leeds client did with her spreadsheet. If you don't measure it, you'll underestimate how much time it saved, or overestimate it.
- Add a second task only once the first one is automatic. Trying to change five habits at once is why most people give up by week three.
- Build in a check step every single time. AI gets names wrong, invents figures, and misreads tone more often than people admit. Read every output before it goes out under your name.
That last point is the one people skip, and it's where the real damage happens.
Where it goes wrong, and it goes wrong more than the case studies say
I've seen a finance report go to a board with a made-up percentage in it because nobody checked the AI summary against the original spreadsheet. I've seen a customer complaint escalate because the AI-drafted reply used a slightly wrong tone that read as dismissive rather than apologetic. I've seen someone submit a client proposal that still had a placeholder company name from the prompt example left in it because they never read it back before sending.
None of that means the tools are bad. It means the people who get the most out of AI at work are the ones who treat every output as a draft from a keen but careless junior, never as a finished piece of work. The people who get burned are the ones who trust it the way they'd trust a colleague with ten years of experience. It doesn't have ten years of experience. It has confidence, which is not the same thing.
The part-time and flexible side of this
One thing that's opened up because of all this is how much AI has made part-time and flexible work more viable, because the admin drag that used to eat a chunk of a short working week is smaller now. Drafting, formatting, first-pass research, all of it takes less time, which means someone doing 20 hours a week can produce close to what a full-time person used to. Worth reading if you're weighing that up, since the real money and time traps of part-time and flexible roles are covered honestly in a separate guide I've written on this site.
If you're bringing outside help in
Plenty of businesses I talk to have got as far as "everyone's using it a bit, unofficially" and want to move to "we have a plan for this." That usually means someone from outside coming in to look at what's already happening informally, work out what's safe, what's risky, and build proper habits and guardrails around it rather than banning it and pushing everyone further underground. If that's where you are, it's worth reading about what an AI consultant costs before you commit to anything, because the range is wider than most people expect and the cheapest option is rarely the one that saves the most time.
Related reading: How to Train Staff to Use AI (Without Wasting a Year on It).
Frequently asked questions
What percentage of employees use AI at work in 2026?
Most workplace surveys now put regular AI use somewhere north of 70 percent of office workers, but a large share of that is unofficial, personal-account use rather than the company's approved tool, which is the part most reports gloss over.
Is AI replacing jobs or just speeding up tasks?
Both, depending on the role. In writing, admin, and parts of customer service, AI has reduced the hours needed for a task, which has cut into paid work for people who charged by the hour rather than by the outcome. In most other roles it's still speeding up tasks rather than removing them.
Why do people hide how much they use AI at work?
Two reasons come up over and over: fear that it makes their previous output look slow or padded, and fear it signals their role could be shrunk or removed. Neither fear is irrational, which is exactly why the hiding happens so often.
What's the biggest mistake people make with AI at work?
Trusting the output without checking it against the source. AI gets figures, names, and tone wrong often enough that treating its draft as final rather than a first pass is where most real workplace mistakes come from.
Related reading: The AI Subscription Stack: What I Pay For (and What I Cancelled) in 2026 and AI Phone Receptionists for Small Business: What They Cost and Where They Quietly Lose You Money.
