An hour saved is only worth what you do with it
The fix for a stretched team is rarely more hours or more hires. It's subtraction: taking the admin a machine should be doing off your people. But decide what the freed hours are for first, or they'll quietly fill up with more of the same.
When a team is stretched, there are two instincts, and they’re always the same: hire someone, or ask everyone to dig in. Both feel responsible. Both are usually wrong, because the problem is rarely too few people or too little effort. The week has silted up with admin, and your people are the ones wading through it.
Asana surveyed more than 10,000 knowledge workers and found they spend about 60% of their time on “work about work”: chasing status, hunting for documents, copying between apps, meeting about the work. Only about a quarter of the day goes on the skilled work they were hired to do. Microsoft’s research lands in the same place. 68% of people say they don’t get enough uninterrupted time to focus, and the average employee spends 57% of the day communicating rather than making anything.
Urgent is loud
There’s a seventy-year-old way of seeing why the week disappears. Eisenhower sorted his work by two questions, is it urgent and is it important, and Stephen Covey later turned that into a simple two-by-two grid that’s now a management staple. The useful insight is that urgent and important aren’t the same thing, and we mix them up constantly.
Look at where the admin lands. The inbox, the status chase, the copy-paste, the report that’s due all feel urgent, because they’re loud and they have deadlines. Almost none of it is important, in the sense of actually moving the business. Meanwhile the important work, the planning, the big customer, the idea that would change things, is rarely urgent, so it never shouts, and it never gets done. A machine can empty the loud box. Every hour it frees could go to the quiet one.
The hour doesn’t stay free
That “could” is doing a lot of work. Time you free up doesn’t stay free. It fills up with more of the same: another round of emails, another meeting, another report someone suddenly has time to ask for. Email and team chat were both supposed to give us time back, and yet here we all are, spending most of the week on work about work.
So before you automate anything, decide what the hours are for. Name the important work that keeps getting pushed back, the plan, the product, the customers you never have time to call. If you don’t give the hour a job, the inbox will.
Find the busywork
Spend a week working out where the hours actually go, because the biggest time-sinks are usually the ones nobody notices. The work that’s ripe for removing has a signature. It’s repetitive, it follows rules, it touches a few tools, and it happens every day or week. Copying enquiries into a spreadsheet. Formatting the same report. Sending the third polite payment reminder. Chasing people for updates.
The cheapest audit there is: for one week, everyone jots down any task that made them think “a robot could do this”. The list arrives fast, because people feel the dull work even when they can’t see the total. McKinsey reckons today’s tools could take on the activities that soak up 60 to 70% of people’s time. You won’t remove all of it, and you don’t need to.
It hides in the small stuff. The work worth deleting rarely looks like one big obvious job. It’s twenty small ones, five minutes here and ten there, spread across the week, which is exactly why nobody fixes it. Write them all down before you judge any of them. The total is always bigger than anyone guessed.
One task at a time
Don’t try to automate the whole business at once. Big “let’s become an AI company” plans stall because they’re too big to finish and too vague to start. One task is neither. Take the worst time-sink on your list, the one everyone groans about, and spend an afternoon getting rid of it. Once it’s running and you’ve felt the hours come back, take the next. A team that removes one piece of dull work a fortnight is unrecognisable in six months, and nobody had to work a minute later to get there.
The first layer is wiring your apps together so nothing gets copied by hand. Zapier and Make connect thousands of apps with simple “when this happens, do that” rules, and neither needs code. Alongside that, let AI do the first drafts. In a controlled MIT study, ChatGPT cut the time it took to write mid-level business documents by 40% and made them better, with the weaker writers gaining the most, and GitHub found developers finished a task 55% faster with an AI assistant. The machine does the first 80%. The person does the part that needs judgement.
The next layer is handing a whole task to an AI agent. You give it a goal and it works across your tools to get there. ChatGPT’s agent mode gets its own browser, and Claude’s connectors reach into Slack, Google Drive, your CRM and the rest. Agents are strong at the doing and still confidently wrong often enough that someone should approve anything that spends money or goes out in public. Treat one like a capable new hire in their first week. Useful from day one, and not left alone with the bank account.
The deepest layer is a custom system that runs the repetitive work end to end, built with something like n8n or Claude Code. That one’s a project, not a setup. It needs looking after and it breaks when your tools change, so prove the value with the first two layers before you build it, and get someone who does this for a living to build it once and hand it over.
Keep people on the work only they can do
The point was never to wring more tasks out of the day. It’s to get your people back to the work a machine can’t do: judgement, relationships, taste, and anything that can’t be undone. The unhappy client who needs someone to actually listen. The hire. The price negotiation. The payment that can’t be clawed back. Automate around those, not through them.
Trust it, then check it anyway. Once an automation looks reliable, people stop checking it, and that’s when a quiet error runs for a month before anyone notices. Keep an eye on anything that moves money or goes out in public, however trustworthy the system has started to look.
And if a “do more with less” plan ever lands on your desk meaning the same work with fewer people, be wary of it. That’s just overwork with a tidier name. The version that works deletes the work, never the people, and then decides what the hours are for.