Insights
Clear thinking on how organisations can adopt AI well. Insights, ideas & tips.
Claude Code and Codex are wired to one model by default, and switching normally means editing JSON, TOML and environment variables by hand. So we built a small menu-bar app that does it in one click. Here is what it does, how it works, and why lowering the cost of switching matters more than it looks.
Everyone's worried about robots taking jobs or a rogue superintelligence. But the genuinely disruptive changes are already happening, more quietly and much closer to home. A lecture by Princeton sociologist Zeynep Tufekci at King's College London made that case compellingly.
People assume an AI advisor must run a vast, exotic stack of tools. Ours is small, it changes constantly, and most of what passes through it gets dropped. We built a live Cowork dashboard to make that honest, and the thinking behind it matters more than the dashboard.
Most people using Claude Cowork are quietly burning tokens they don't need to. This is a practical guide to the habits, settings, and advanced moves that make a real difference, without turning AI into a full-time job to manage.
A relaxed walk through the stack and workflow behind futureformed.io. Claude, Cowork, GitHub, Vercel, Buttondown, Substack and a Next.js codebase wired together so the site can be updated in plain English. Honest on the trade-offs and on when DIY is the wrong call.
Most of us have no idea how much energy our AI usage consumes. AI Wattch is a browser extension that makes that invisible impact visible, in real time, without sending your data anywhere. It is a small tool with a genuinely useful nudge built in.
AI chatbots were the first budget line most organisations noticed. Agents will be the next, and this one will be harder to manage. When agentic workflows spread across a business, cost stops being a fixed subscription and becomes something you have to actively govern.
Most AI tools are built for capability, not efficiency. GreenPT is a rare exception: a purpose-built AI platform that runs on renewable energy, uses models optimised to consume 20-30% less energy than standard LLMs, and is honest about its footprint. It deserves more attention than it's getting.
Most businesses have a continuity plan for power outages, key people leaving, and server failures. But how many have a plan for the day their preferred AI provider goes down, hikes prices, or changes direction? That gap is becoming a real problem.
Running too many AI pilots with no clear success criteria is quietly killing your AI strategy. Here's why you should run multiple cheap experiments, but kill them fast and hard if they don't deliver.
Thinking about migrating from ChatGPT to Claude? It takes about 30 to 60 minutes, spread across a day or two. This is a step-by-step guide to exporting your history, extracting your context, and importing it into Claude so you pick up where you left off.
Every organisation is buying AI tools. Far fewer are thinking about who (or what) is going to coordinate them. The AI orchestrator is the most important role in modern AI adoption, and most businesses haven't got one.
Most businesses focus on which AI tools to buy and how much they'll save. But the real question, what would your people do with their time if AI handled the boring stuff, is being left unanswered. The time dividend is real, but only valuable if you're deliberate about where it goes.
Handing out AI subscriptions isn't AI transformation. It's procurement. The distinction matters enormously. AI is reshaping how decisions get made and how work gets done, and the companies that treat it as an app to bolt on will end up exactly where they started.
The UK government's new AI Skills Hub offers hundreds of learning modules, but the sheer volume can be paralysing. The real key to AI education is simpler than you think: start by using it, then find structured support to go further.
By the time leadership has decided it's time to think seriously about AI, the staff have already started. Studies put the proportion of employees using unsanctioned AI tools at 20–30%. The instinct is to treat this as a governance problem. That instinct misses the signal.
Some AI pilots succeed. Most AI programmes don't. The gap between them is rarely in the technology. It's in whether the people who need to use it actually do. This is a change management failure, and it is remarkably common.
When someone comes to us for help with their AI strategy, the first question we ask is a simple one: what problem are you trying to solve? The answer tells us almost everything we need to know, including, sometimes, that AI isn't the right answer at all.
Stay sharp
Occasional, considered articles on AI adoption and transformation. No noise. Unsubscribe any time.
AI transparency: The articles on this page are written by Liam. Each post includes its own disclosure where relevant.