Millwright

One workflow at a time, with your people in the loop.

Working with us follows five stages. You can stop after any of them. Nothing is built until you've seen, in writing, what it is, what it costs and how we'll know whether it worked.

The five stages

  1. A conversation

    We start by listening. How does the business run, what work repeats, what tools do you already use and what would you like to be different? It's free and there's no preparation needed.

    You bring: A rough picture of your week, and one or two tasks that annoy you.

    You get: An honest view of whether AI is likely to help, and where.

  2. A workflow review

    We look closely at the most promising pieces of work. Who does each step, what information goes in, where it waits, what goes wrong and how often.

    You bring: Time with the people who do the work, and a few real examples of it.

    You get: A short list of opportunities, each with what it would take and what could go wrong.

  3. A proposal

    We write down one workflow to start with: what we'll build, what your team will do, how we'll measure the result, how your data will be handled, the cost and the timing. All of it agreed before work begins.

    You bring: A decision-maker, and your questions.

    You get: A plain document you can say yes or no to. No surprises later.

  4. A pilot or implementation

    We build the workflow alongside your team, fitted around the tools you already use. People review what the AI produces, especially at the start, and we adjust as real work passes through.

    You bring: Someone who owns the workflow day to day, and feedback as you go.

    You get: A working workflow on real tasks, with review steps where they matter.

  5. Evaluation and continuing support

    We compare the result with the baseline we agreed. Then you decide: keep it, improve it, extend it to the next workflow, or stop. We stay available to maintain and adjust what we built.

    You bring: Your honest verdict.

    You get: Measured results, in your own numbers, and a clear next step.

How we'll know it worked.

Before the pilot starts, we note how the work goes today. That's the baseline. Afterwards we measure the same things again. These are the usual four, and we'll add whatever matters most to your business.

  • Time per task

    How long one request, quote or record takes from start to finish.

  • Completeness

    How often everything needed is there the first time.

  • Corrections

    How often a person has to fix what was prepared.

  • Handoff delays

    How long work sits waiting between one person and the next.

Questions owners ask us.

If yours isn't here, ask it in the first conversation.

Where should a business like ours start with AI?

With one piece of work that is frequent, repetitive and a bit annoying, where a mistake is easy to catch. Answering routine inquiries, preparing quotes and checking incoming paperwork are common first choices.

We'd steer you away from starting with anything where an error is costly or hard to notice. That can come later, once you trust the approach.

How do you work out which opportunities are worth it?

In the workflow review we ask how often the task happens, how long it takes, where it waits and how often it goes wrong. Then we weigh that against how hard it would be to build and how much checking it would need.

The best first project is rarely the most exciting one. It's the one with a clear before and after.

Will it work with the tools we already use?

That's the aim. We'd rather fit around your inbox, your spreadsheets and your booking or stock system than ask you to adopt something new. What can be connected depends on the tools, and we check that during the review, before we propose anything.

What does our team need to do?

Show us how the work is done today, give us a few real examples, and review what the AI produces during the pilot. One person should own the workflow on your side. We'll keep the time we ask for small and predictable.

Who's in charge once it's running, the AI or us?

You are. We design each workflow so that people approve what matters: prices, promises, anything sensitive, anything going out in your name. Where the AI isn't sure, it should say so and hand over, not guess.

The goal is to take routine work off your team's plate so they have more time for the parts that need them.

What does it cost, and how long does it take?

Both depend on scope: how many steps the workflow has, what it needs to connect to and how much review it requires. The first conversation is free. After the workflow review you get a proposal with the cost and timing written down, and nothing starts until you've agreed to it.

What if it gets something wrong?

It will, sometimes, which is why review is part of the design and not an afterthought. We agree up front what gets checked by a person, how mistakes are reported and corrected, and what the workflow should do when it can't complete a task. Corrections are one of the things we measure.

What happens to our data?

Before any work starts we agree in writing what information the workflow needs, where it's processed and stored, who can see it and how long it's kept. We use the least we can. There's more on the data handling page.

What happens after the first project?

We look at the results together. If it worked, you might extend it or pick the next workflow. If it didn't, we say so and work out why. Either way you're not locked in, and we stay available to maintain what we've built.

Start with the free conversation.

No preparation needed. Bring a rough picture of your week and one task you'd happily never do again.