Your people aren’t against AI. They’re waiting for proof.

I coach teams and individuals from AI-hesitant to AI-productive, one small, safe task at a time — and build the small tools that make the first step easy.

Book a 20-minute call How it works

A non-coder who built 46 working software projects with AI in two years — after 34 years coaching people through change.

Monday receipt-chasing25 emails, same three sentences, two hours a week
Define
Shadow
Refine
Working
You
Writing all 25 yourself
AI assistant
—drafts you’d send untouched

Pick one small, repetitive job nobody would miss.

The assistant runs alongside you and is relied on for nothing — until the evidence says otherwise.

The licences are bought. The usage is flat.

Most people aren’t resisting AI. They’re sensible people waiting for evidence that it works on their job, not someone else’s.

Training courses don’t give them that evidence. Demos of other people’s work don’t either. What changes minds is watching an AI assistant do one real piece of their work, alongside them, until they can see for themselves that it’s right.

Nobody is asked to trust it first. Nobody is asked to stop doing things the way that works. The evidence does the persuading.

Start with an intern, not a revolution.

Imagine you’ve been given an intern. In week one they know nothing, and training them is a pain. A month later they work the way you like, at your pace — and you’re still the one who signs things off.

That’s the whole idea. You stay the senior. Early mistakes are expected, the way they are with any new starter. And the intern’s work gets checked before it goes anywhere, until it has earned otherwise.

Every professional has trained someone. No technical vocabulary needed.

  1. Define

    Find the lowest-risk, most repetitive task in your week — something small nobody would grieve — and agree what a good result looks like before anything is built.

  2. Shadow

    The assistant does the same task alongside you. You keep doing it your way. Nothing depends on it yet.

  3. Refine

    The gap between your version and its version is the lesson. Each correction becomes a standing instruction, in plain English.

  4. Working

    When it matches your standard consistently, you hand it over — and pick the next small task. Trust is earned one task at a time.

Five rungs. Where are you now?

Most people stop at the first rung and conclude AI is a clever search box. Each rung up hands over a little more — and each one is earned the same way. Choose the one that sounds like you.

It works on people who never meant to try it.

A non-coder who built a software estate

Two years ago I couldn’t write a line of code. I learned AI the same way I now teach it: one small task, shadowed until it earned trust, then the next. My first product tried to do everything at once and taught me exactly why that fails.

Today I run a portfolio of live products — price comparison sites, a coaching platform, a community flood-warning system — with AI agents working from a shared ticket list, checked at every step. The method on this page is the one I use every day.

46software projects
~900klines of code
0written by hand

A flood warning a village relies on

An early-warning site for a flood-prone Oxfordshire village, reading live river gauges upstream. Built and run with AI, read by residents when the water rises.

See the live site

From their eighties to the engineering floor

Retirees in their eighties, personal-training clients, small-business owners and sceptical senior developers — coached one small task at a time. The steps don’t change. The first task, the words and the pace do.

Never chased, always volunteered

I don’t push AI on anyone. People come to it when they see something useful working on a job like theirs — and then they bring the next person.

The coach gets a console. You get a link.

I built a coaching console to run the method. Each person gets one private web link — no install, no login — showing their own tasks moving from Define to Working, with the evidence on every card.

Mapped

Supplier statement matchingAgreed: flags any mismatch over £1

Shadowing

Month-end summaryWeek 2 of 3

In service

Receipt-chasing drafts24 of 25 sent untouched

Parked

Invoice codingTried, shelved — no shame
  • Minimum data by design. No surnames, no email addresses, no manager field — it isn’t hidden, it doesn’t exist.
  • Nothing held in secret. People can read and annotate everything recorded about them.
  • Coach, not auditor. Leadership sees organisation-wide patterns, never an individual’s progress.
Open an example board

Fictional example: an accounts professional’s first three weeks.

For teams on large, old codebases.

The hard part of AI at team scale isn’t the AI. It’s people and agents trampling each other’s work, and sessions losing the thread halfway through a hundred-thousand-line problem. Both are discipline problems, and both are solvable.

One list of who’s doing what

Every piece of work is a ticket with a single owner. Two agents — or two people — can’t pick up the same job, so nobody overwrites anybody.

Evidence, not long memories

Each AI session hands the next a short written record of what it found. Fresh sessions start from facts, not from a sprawling chat history.

People approve, not relay

Agents do the work directly and show their evidence. People review and sign off, instead of copying and pasting between tools.

Two teams, same work — then twice as many teams.

An AI-assisted team builds a low-stakes internal job while a paired team reviews every step. Every difference between them is evidence. At the end of the cycle, the reviewing team becomes the next building team — so capability doubles each round instead of waiting on a training course.

A live board shows work in progress and warns when two people are about to change the same files — before they collide. It tracks work, never workers.

Ticket 214 · Anna Ticket 219 · Agent Both touch billing.js

Ways to work together.

Individuals and small groups

One-to-one coaching

Short, confidential sessions around one real task from your week. Open to anyone who volunteers — nobody is sent.

4 to 8 weeks

A team pilot

One low-stakes internal job, delivered by an AI-assisted team with a paired team reviewing. You finish with a working result and a second team ready to lead the next one.

Days, not months

Small tool builds

A small working tool that removes one named frustration for one named team. The quickest way to show a sceptic what’s possible.

If you want a course library, a compliance tick-box or AI imposed on people who didn’t ask for it, I’m not the right person.

Fear is a coaching problem, not a technology problem.

Ewan Norrie smiling while presenting a talk

I’ve spent 34 years helping people change how they work and live — as a personal trainer, sports therapist and psychotherapist. Then, two years ago, I started building software with AI from a standing start.

That combination is the point. I know what it’s like to be the person who doesn’t understand the technology, because I was that person. And I know that people don’t change because they’ve been told to. They change when it’s safe to try, when they stay in charge, and when the evidence is theirs.

Based in Oxfordshire. Working across the UK, in person and remotely.

Start with one small task.

Tell me what your team does most often and likes least. In twenty minutes we’ll find the first thing worth handing across — and what good would look like.

Book a 20-minute call hello@ewannorrie.com