Software should feel like a quiet colleague.

Most companies don't need more software. They need fewer places to look. I'm a designer who also builds, and I think about tools the way I think about rooms and brands: what people need to notice, what they can ignore, and how the working day should feel.

The cheapest time to change your mind is now.

Every decision gets more expensive as it hardens into code, data and habits. So I make the future clickable in days, put it in front of the people who will use it, and only then build.

An illustration of the cost-of-change curve, not measured data.

Six ideas behind every tool I build.

I design and I build, so these aren't slogans. They are the rules I follow when I decide what software should do, and what it should leave to people.

  • Automate remembering, not judgment — Templates, checklists, dashboards and AI should handle retrieval, formatting and routine checks. Interpretation, taste, trade-offs and relationships stay with a person. — Your team stops remembering and starts noticing.
  • A dashboard must tell the truth — Every status is defined by evidence, not optimism. "Almost finished" becomes states you can see. Red is not failure when it appears early enough to help. — You see problems while you can still fix them.
  • Fix the condition, not the person — If the same mistake happens three times, the system needs a review. Instead of asking why someone was careless, ask why the error was easy to make and hard to detect. — Fewer warnings, fewer mistakes, less blame.
  • Design for the tired user — Real users arrive distracted, in sunlight, after a mistake. Empty screens, exceptions and undo are designed on purpose, not left to chance. — People can recover without calling anyone.
  • Not all friction is waste — Before automating a step, I sort it: waste, protection, training, or unclear. Bad friction goes; the friction that protects quality stays. — Faster where it should be, careful where it must be.
  • Ship the smallest honest version — One scoped pilot that solves one real problem. A draft a person reviews, rather than a system pretending to be autonomous. — Something useful in weeks, and real evidence for the next step.

    • Keep human. Co-pilot. Automate.

    Before any AI or automation, every task in the workflow goes into one of three columns. Most of the value is in being honest about the first one.

  • Keep human — Interpretation, taste, trade-offs, relationships, anything with consequences. — A person decides. The system makes their information clear.
  • Co-pilot — Drafts, comparisons, summaries, first versions. — The machine proposes; a person checks. Work is labelled generated, verified or approved.
  • Automate — Backups, reminders, formatting, conversions, routine checks. — Runs on its own, so nobody has to remember it.

    • Every extra choice costs time.

    The more options on a screen, the longer people take to decide, and the more mistakes they make. A good tool shows the next likely step and hides the rest until it's needed.

    An illustration of Hick's law: decision time grows with the number of choices, quickly at first and then more slowly.

    Designed for tired people on a busy day.

    Nobody uses work software at their best. These are the principles I check every screen against.

  • Hick's law — Decision time grows with the number of choices. — One main action per screen; the rest one tap away.
  • Recognition over recall — People recognise far more easily than they remember. — Show the options, the last values and the status, never make people type from memory.
  • Progressive disclosure — Too much at once overwhelms; too little hides what matters. — Simple first, details when asked for.
  • The default effect — Most people keep whatever is already chosen. — Safe, sensible defaults, so the easy path is also the right one.
  • The Zeigarnik effect — Unfinished tasks stay on our minds. — A clear list of what's left today, and a real sense of done.
  • Fitts's law — Big, close targets are faster and safer to hit. — Large buttons where hands and eyes already are, especially on phones.

    • Adapt to the tool, or the tool adapts to you.

    Big platforms promise everything and ask your team to change how it works. I'd rather start small, with exactly what your people do.

  • Months of setup before anyone uses it. → A first working version in weeks, used every day from then on.
  • Hundreds of features, most of them never opened. → The few things your team does every day, done very well.
  • Stops when the internet does. → Works offline, and backs itself up on its own.
  • AI added because it sells. → Automation only where it saves real hours, and nothing that pretends to be magic.

    • Automation is most valuable when it removes repeated remembering while leaving judgment with the person.

    Automation is most valuable when it removes repeated remembering while leaving judgment with the person.

    How I work with people.

    Good work is also a promise kept. These are the standards I hold myself to on every project.

  • The minimum reliable promise — I would rather make a smaller promise I can keep than a bigger one I can't support. Every date comes with its conditions: what must be approved, and by when. —
  • Facts before reassurance — What is confirmed, what is still pending, and when the next update comes. If there is bad news, you hear it early, and you hear it from me. —
  • Calm under pressure — I absorb urgency instead of passing it on. Problems come to me early, because people know they will be handled, not dramatised. —
  • Decisions in writing — After every important conversation, a short note: the decision, the owner, the deadline and what is still open. Clear to someone who wasn't in the room. —

    • How I would work with you.

    Small, real and in use quickly, then better every week.

  • Sit with the team — A few days next to the people who do the work, watching where time goes. — a map of the real workflow
  • Find the mess — Where data lives twice, where people wait, where mistakes happen. — the three problems worth solving first
  • Prototype — A clickable version in days, tested with the people who will use it. — a decision made on evidence
  • Build the smallest real thing — One tool that solves those problems, in use as soon as possible. — software your team opens every morning
  • Automate what repeats — Backups, reports, conversions and reminders, once the work is clear. — hours back, every week

    • Running on spreadsheets?

    Tell me how the work moves today, and where it gets stuck.

    Start a project