Ask this after someone buys: what almost stopped you? Not “Why did you buy?” Not “Any feedback?” “What almost stopped you?” is the one that tells you where your conversion friction actually lives. Because “why did you buy?” gets you the polite, post-rationalised version.

Your AI policy is not in anyone's browser history. That's the uncomfortable truth behind most AI governance efforts. You can publish a PDF, run a lunch-and-learn, even make people click "I agree". But when someone is under pressure, they use whatever saves them time. And if

If your offer can be bought, AI will help people buy it faster. If your offer must be sold, AI will mostly make the repetitive parts cheaper. That sounds like a throwaway line. It isn’t. It’s the difference between: - AI increasing your top line (more demand, faster conversion)

The UK has crossed a line. Welfare spending: £333 billion Income tax revenue: £331 billion We’re now paying out more than we bring in from workers. This is unsustainable.

Cash leaks kill more startups than bad products. I heard a slightly brutal reminder in a talk on founder money mistakes: when you start making a bit of money, the “messiah consultants” show up. They promise to take you to the next level. They sound confident. They have a price

Debugging is the skill you need when the AI is wrong. If AI use erodes the spot-the-mistake muscle, you end up with a team that can ship quickly and fix slowly. That is not leverage. This is the bit that’s easy to miss if you’re a founder watching velocity charts. AI makes it

Security used to be a specialist sport. AI is turning it into a volume sport. Anthropic just announced Project Glasswing: a group of big tech and security firms using an unreleased model to find and fix serious vulnerabilities at scale. The detail that hit me: they describe

If AI can change customer records, send emails, approve refunds, or alter a price, ask yourself: Who can tell what it changed, when, and why? Because that question is the line between “useful automation” and “we’ve just built an audit nightmare that types quickly”. Right now,

Instead of asking for a full build quote, buy a small, paid discovery (a short planning sprint) with clear outputs. Treat it like an option contract: you pay a sensible amount to find the unknowns, then you decide whether to exercise the option and build. Most founders and ops

Security has a new problem: attackers can now scale curiosity. That sounds abstract, but it’s brutally practical. For years, vulnerability discovery was constrained by specialist time. You needed the right people, the right skills, and enough hours to stare at code and systems

If your portal looks inconsistent, your team is paying an invisible tax every day. You see it as: - slower delivery (because nothing is reusable) - more rework (because decisions get reversed late) - longer meetings (because everyone debates the same stuff) - that creeping sense

One of the biggest issues facing the UK right now

Federer won about 54% of points across his career. If your last project feels like you’re losing points every day, that number should calm you down. Because most leaders I speak to are quietly carrying this belief that good teams win by being right all the time. They don’t.

In the 2025–26 financial year, the UK Treasury received £331 billion in Income Tax. In the same period, the Government spent £333 billion on welfare. This situation is clearly unsustainable and requires urgent change. For the first time ever, we’re now spending more on welfare

In 2026, "we added AI" will sound like "we made it mobile friendly". Not because AI is useless. Because AI is becoming infrastructure. That shift matters if you are a founder, CEO, or ops lead making product and platform decisions. When a capability becomes normal, it stops

GOV.UK is the best B2B portal I’ve ever used (and it isn’t trying to sell me anything). It has a line most internal tools should steal: “Simpler, clearer, faster.” What makes it good isn’t the colour palette or some trendy UI kit. It’s the way it respects your

Your biggest operational risk is a double click. I spotted a tiny line in the Readwise changelog: they fixed a bug where quickly double-tapping the signup/login button could trigger an error page instead of ignoring the duplicate tap. That sounds like minor UX polish. It isn’t.

AI coding tools are like wrist straps in the gym: they let you lift heavier, but they can hide weak grip. I love these tools. We use them. We will keep using them. But if you are leading a team, there is a failure mode that does not show up in your sprint report. The team

Stop calling it "user research" and start calling it "expensive assumption removal". Because that is what you are really doing. Most teams do interviews like this: - "Would you use an app that did X?" - "Which features would you want?" - "How much would you pay?" And then they

Profit comes from the loop, not the feature. A competitor can copy a screen. They can even copy a workflow. What they struggle to copy (and what most leadership teams underinvest in) is the system that keeps improving while you sleep. Because features are visible. Loops are