June 29, 2026 · by David Gilbert · 3 min read · Tech & AI
Between repairs, web projects, radio, and whatever 3D-printed emergency someone's brought in that week, my admin pile is constant: quotes, invoices, follow-up emails, social posts, supplier emails I keep meaning to answer. So last month I set myself a rule — for thirty days, before I touched any admin task myself, I'd try handing it to an AI tool first and see what actually held up.
What Stuck
The single biggest win was drafting. Quotes, follow-up emails, the first pass of a blog post like this one — having something to react to and edit is so much faster than staring at a blank page. I'd give it the bones of what I needed to say and it would hand back a structured draft in seconds. I still rewrite plenty of it in my own voice, but the blank-page problem basically disappeared.
Second was summarising. Long supplier email chains, meeting notes from a client call, even transcript dumps from podcast recordings for the radio show — having those condensed into three bullet points instead of three paragraphs saved real time, every single day, not just occasionally.
Third, and the one that surprised me most, was using it as a sanity check before sending anything client-facing. A quick "does this sound reasonable, am I missing anything obvious" pass caught a couple of genuinely embarrassing oversights before they went out the door.
What Didn't Stick
Anything that needed real judgement about a specific person or situation fell flat. AI doesn't know that a particular client always pays late and needs a gentler nudge than the template suggests, or that another one prefers blunt and brief over polite and padded. I ended up doing those by hand anyway, which tells you something about where the actual value of "knowing your customers" sits.
It was also unconvincing on anything genuinely technical and specific — diagnosing an actual fault from a vague description, for instance. General troubleshooting logic, sure. The kind of pattern-matching you build from genuinely having seen ten thousand broken phones in person, not so much.
This Tracks With What I'm Seeing Everywhere Else
I'm not the only small business owner running this kind of experiment right now — depending which survey you read, the typical small business is juggling something like half a dozen different AI tools across content, admin, and customer-facing work, and most of them say they're planning to add more rather than cut back. The pattern I keep hearing echoed by other business owners matches mine almost exactly: the time savings show up in drafting and summarising, not in judgement calls or anything that needs real context about a specific customer.
So What's Actually Changed, Long-Term?
I haven't kept every habit from that month, but two have genuinely stuck: I draft-first instead of blank-page-first for anything written, and I run a "what am I missing" check before anything client-facing goes out. Neither replaced my judgement. Both gave me back time I used to lose to staring at a cursor.
If you're a small business owner wondering whether it's worth the fuss, my honest take after a month of forcing myself to use it is: yes, for the boring repetitive stuff, no, for anything that needs you to actually know your customer. Which, conveniently, is most of what makes a small business feel like a small business in the first place.