I read this piece from the BBC, and Sabine Zetteler’s words — “Why would I bother to read something someone couldn’t be bothered to write” — really resonated with me. I’ve long felt that when AI is used to generate content, it often becomes effortless and as a result, thoughtless and empty.

Despite my back catalogue, talks, and elements of my job, I’m actually something of an AI skeptic. It’s interesting, and at times an amazing, piece of technology, and it’s certainly ‘here’. Whether it’s here to stay, or whether the big players will turn off their free tiers as a cash grab after pulling up the ladder once they finish consuming the sum total of everything humanity has ever created — well, time will tell won’t it.

I have, and continue to, criticise those who utter phrases like “I’m not a computer person” as a way of dismissing basic computer literacy and inefficient practices. If AI is here to stay, it’s worth having a proficient understanding of it. In its current form, VSCode plugins like Copilot, services like Cursor, or even just the helpful AI elements of Office365 for meeting summarisation or writing assistance in Outlook and Word. There’s plenty to make use of in daily workflows to make our lives easier.

I feel reasonably safe discussing this topic — I’m not new to AI, nor am I just on the periphery of the technology. I work with (and for) an organisation that has literally won awards for our implementations, and I’ve spoken about AI at events like Women in Tech York and most recently, at Leeds Digital Festival. I had attendees hack together their own RAG and knowledge management starter

With things like ChatGPT being possibly the most common AI tool for most people, there’s a lot of noise around writing ’the perfect prompt’ and ‘prompt engineering’. But as long as you’re reasonable about what you want and how you want it, the responses are pretty solid.

Not asking for everything all at once and multi-shot prompting and piecing units of work back from your AI seems to result in better quality and ensures you still understand the thing you’re building/writing/designing. Compare that to just asking for “a Noughts and Crosses game written in VueJS” and YOLO-committing it to GitHub and sharing it on Hackernews for the points.

AI code is legacy code from day one – Krrish

There are a few named frameworks for writing a good prompt. Like the old adage of the compliment sandwich, they can be used well or used poorly, its just guidance, not gospel. My favourite is the COSTAR one, mostly because I can remember it (and because the name is close to COPILOT):

• Context – Give background. It’s looking for likely contextual matches, so help narrow the pool.
• Objective – What are you trying to achieve?
• Style & Tone – For me: UK English, no hallucinations, and thoughtful responses.
• Audience – Who’s this for? Often framed as “Act as a…” or “You are in the role of…”
• Response – Output format: code, text, table, whatever.

Our tone, humour, and judgment are what make blog posts worth reading. Writing AI-driven drivel for other AI or crawlers to read for SEO just builds on the death of the internet — replacing what we’re great at with bland, homogenous ‘Content’.

That doesn’t mean AI can’t be part of the process, it can help us become more competent and accessible to wider audiences, without turning into soulless slop merchants. Back to Sabine’s point of ‘if I couldn’t be bothered to write this, why should I expect anyone to read it?’, equally if its written badly nobody is going to read it either.

But if I have written it — and I’ve written it badly — wouldn’t it be helpful if something trained on how language fits together could suggest edits? I don’t want AI to write for me. I don’t want it to act as the pilot, I want it as a co-pilot, an editor, and a proofreader.

I fully support experimenting with the AI tools at our disposal - especially while the major players are handing out free access left, right, and centre. But we should do so without surrendering our agency. We should absolutely be looking at our workflows and offloading busywork. Thinking and creativity should remain the domain of beings that can actually think, not just regurgitate - until AGI is a thing anyway, then I’ll update this.

Use the toys, let them aid but not drive. Let them proofread, draft, reformat, summarise, nudge, or restructure. But don’t let them replace your curiosity, your voice, or your thinking. That’s the bit thats probably worth reading.

Anyway, please subscribe and read my upcoming posts:

• Unlocking the Power of AI to Empower Innovation at Scale
• How I Got 300% More Efficient By Delegating Thinking to a Bot
• 5 Ways AI is Already Replacing Your Coworkers (And Why That’s a Good Thing)