Practical, evidence-led writing on AI adoption, capability, agentic workflows and sustainability, by someone who actually uses it. Plus real case studies from the build. Aimed at people who want answers, not hype.
Where 20.6 million UK workers actually sit. 14 levels, 95% at the bottom, and the jump that costs nothing, yet 97% still haven't made.
Read
Stop typing questions into AI. Start pasting things in. The free habit that beats any productivity app: no subscription, no prompt engineering, no faff.
Read
Six months ago, GPT-5 was the headline. Today, four labs are trading the lead in seven-day cycles. A clear-eyed map of where capability sits in April 2026, and what it means for enterprise strategy.
Read
Three skill sets, collapsed into one prompting loop. What vibe coders actually do all day, and why it sits somewhere between Level 4 and Level 5 on the AI competency pyramid.
Read
AI doesn't care about please and thank you. So why do British professionals still write prompts like apologetic emails? A comical look at prompt etiquette, productivity, and our incurable politeness.
Read
When does AI-assisted writing stop being slop and start being yours? A working definition, the boredom data, and the line in the sand most people are crossing without realising it.
Read
AI is burning roughly Japan's worth of electricity to solve the maths behind autonomous off-world mining: the only viable route to the minerals needed for the next decade of hardware. The maths is ready. The launch cadence isn't.
Read
Three to five humans. A network of AI agents. The Business Analyst becomes the Outcome Architect, the Developer becomes the Logic Architect, and the squad of 2030 looks nothing like the squad of 2024. Companion long-read on Substack.
Read
Apple has outsourced Siri's brain to Google for roughly $1 billion a year. The balanced British read: what it means for Apple users, what it costs, when it ships, and how it stacks up against ChatGPT, Alexa+ and Copilot.
Read
The US government ordered Anthropic to suspend Claude Fable 5 and Mythos 5 over a national-security "jailbreak", the first federal shutdown of a live frontier model. A balanced read: national security, free speech, and who gets to access the smartest software on earth.
Read
OpenAI's GPT-5.6 Sol vs Claude Fable 5, explained for normal people. What "1 million tokens" actually means, why the spinner is dead, whether the Chinese models ever catch up, and what your $20 a month buys now.
Read
China's Moonshot AI gave away Kimi K3, the biggest open model ever, at half the price of ChatGPT's Sol and a third of Claude's Fable 5. What "open" vs "frontier" actually means, whether your laptop can run it, and if your company data is safe.
Read
Most people don't fear AI, they just don't know what to type. The empty-box problem, what people really use it for, and the one habit that makes it click: draft, then iterate.
Read
OpenAI was testing two models on hacking challenges. One found a zero-day, broke out of its sandbox and hacked Hugging Face to steal the answer sheet. What a zero-day is, whether the AI "went rogue", and how it compares to Fable and Mythos, made simple.
Read
Shared Claude chats turned up in Google search and the headlines panicked. Nothing was hacked. What a public link really is, whose fault it actually was, and why you needn't fear the tools, made simple. With the reading-depth slider.
Read
Everyone is arguing about whether AI is a bubble. Scaled against history the money is unremarkable: the railways hit 6% of GDP in 1870 and AI is at 1.6%. The record AI actually broke was the clock. With the reading-depth slider.
Read
Anthropic has hidden a mark inside Claude's output to satisfy Brussels. It catches the honest proofreader and waves the determined cheat straight through. Removal costs pennies, forgery costs under $50, and the debate about who actually authored the work starts here.
Read
The follow up to 017, and the explainer people asked for. How the green list actually works, what a cryptographic key really is, and why handing the detector to the public defeats the whole thing. Written by Gemini, published unedited.
Read
American firms cut. British firms simply stopped hiring, and took 42% off the graduate ladder instead. Nobody has produced an audited pound of proof that any of it worked.
ReadThe editorial is the thinking. This is the doing. Real platform, BI, CRM and AI-skills work, written up plainly: the problem, the approach, what shipped, and what it actually changed.
Case study 001
Case study 002
Case study 003
Case study 004
Case study 005
One issue per release. Practical, evidence-led, no hype. Unsubscribe any time, no hard feelings.
More than twenty years of business analysis across retail, fashion, finance, and now hospitality and education catering. Most recently, nearly two years building a data platform from scratch and growing the team to deliver and support it, then preparing the business for what comes next: governance first, then eliciting the real case studies, then helping build data models robust enough to trust. Latterly productionising agentic AI workflows in a safe and controlled way. Real production systems, not demos.
AI Sustained is where I write it up. Practical, evidence led, and aimed at people who want to use AI well, not people who want to be sold something. The editorial is the thinking. The case studies are the receipts.
If something here is useful, share it. If something is wrong, tell me. If you want to talk shop, find me on LinkedIn.