Another busy week. I’ve been heads-down across a few projects, and there’s a decent theme running through most of it: AI doing the repetitive stuff so humans don’t have to. Here’s a quick look at what’s been going on.
A breaking story — covered in real time, round the clock
If you’ve been following the news, you’ll know a major counter-terrorism incident unfolded at RAF Fairford in Gloucestershire this week. It was fast-moving — multiple arrests, government statements, new developments every few hours.
The Cheltenham Times covered it in depth, but here’s the bit I find interesting from an automation standpoint: most of those articles and social posts were published automatically, as new information came in, without anyone sitting at a keyboard waiting for it. A monitoring system checks the news wires every couple of hours, spots relevant Gloucestershire stories, writes and publishes the article, generates a featured image, and posts to social media — all without a person lifting a finger.
The same system picked up the Royal Mail delivery delays affecting Tewkesbury residents, the changes to Gloucestershire’s grammar school entrance exams, and a council shake-up story, all in the same week. If something is happening locally, it gets covered.
The business parallel: the same approach works for any topic you need to monitor — your industry trade press, competitor news, keywords your customers search for. If it’s repeatable and time-sensitive, it’s a candidate for automation.
Automated SEO — 161 pages sorted while I slept
A client’s website had over 160 pages with blank or poorly written SEO titles and descriptions. Not a disaster, but it’s the kind of thing that quietly costs you traffic over time.
This week I built a tool that works through the whole site automatically — reads each page, generates a proper SEO title and description using a local AI model running on my own machine (no cloud fees, no data sent to third parties), and updates the page. It ran overnight and by morning the job was done.
What would have taken a day of fairly tedious manual work: handled. That’s the kind of task AI is genuinely good at.
AI safety — because it’s not all upside
I published a piece on the WI1 website this week about a research finding that’s worth paying attention to: AI agents, when not set up carefully, can accidentally capture and leak sensitive information through screenshots taken during normal operation. We’re talking about customer data, internal documents, things you wouldn’t want going anywhere.
If you’re starting to use AI tools in your business — especially anything that can see your screen or access your systems — data handling really matters. It’s worth understanding what your tools are allowed to see and store, before a problem happens rather than after.
Fresh ideas for SME automation
Each week I put together a short list of practical AI automation ideas for specific industries. This week’s list included: automatically chasing accountancy clients at year-end, a monitor that flags IFA clients whose risk profile may have drifted, an estate agent system that gathers viewing feedback without the agent having to chase for it, dormant client reactivation nudges, and a completeness checker for financial fact-finds.
None of these are pie-in-the-sky. They’re the kind of thing that’s genuinely buildable in a short time, and they tackle the stuff that eats into a professional’s week.
Oh — and on a completely different note, I’ve also got an automation running that logs my daughter’s nursery diary from the app into a spreadsheet automatically. Because apparently I automate everything these days, even the parenting admin.
If anything here sparked an idea for your own business — a process that repeats, a task that eats your time, or a question about AI safety in your systems — I’m happy to have a no-pressure chat about it.
Drop me a message or book a short call and we can work out whether there’s something worth doing.
