How We Work.

Every project runs through the same five stages: initial meeting, audit, proposal, build, handover. Each one is fixed-fee or free, and each has a clear exit if the numbers don’t stack up.

Five steps.
Nothing fancy.

  1. The Initial Meeting

    A free chat to figure out what you actually need, before anything is scoped or billed. If you already know what's broken, we go straight to a proposal.

    ≈ 30 minutesFree

  2. Path B starts here

    The Audit

    We spend real time with your team. Watch the work happen. Find the slow bits, the tedious bits, the bits where humans are doing what machines should, and the bits where the opposite is true. You get a written brief at the end, with or without us.

    An hour to a monthFixed fee

  3. Path A starts here

    The Proposal

    We pitch only the things that will actually pay back. Each one comes with a baseline measurement, a target, and an honest cost. If a problem isn't worth automating, or if a simpler tool already exists, we say so, in writing.

    ≈ 1 weekFree

  4. The Build

    We build, test, and measure against the baseline. You see progress weekly. If it doesn't beat the baseline by the target we agreed on, it doesn't ship, and you don't pay for the part that didn't work. That's the deal.

    Project-dependentMilestone billing

  5. The Handover

    You get working systems, the data, the documentation, and a person you can call. No lock-in, no proprietary middleware, no "ongoing managed service" you can't escape. The whole point is that this keeps working after we're gone.

    PermanentYours to keep

The rules. What we do, what we don’t.

The five steps are the process. These are the rules we hold to while running it, so you can check us against them.

“If AI isn’t moving a real number on a real spreadsheet, we’re not building it.”The first rule
◆ What we do
  • Measure a real baseline before building anything.
  • Only bill for builds that beat the target we agreed on.
  • Build with open tools you can swap out later.
  • Tell you, in writing, when AI is the wrong tool.
  • Document everything in plain English.
  • Answer when something breaks.
◆ What we don’t
  • Build demos that look great and do nothing.
  • Sell "AI transformation" programs.
  • Hide behind jargon.
  • Charge retainers for systems that run themselves.
  • Lock you into proprietary middleware.
  • Take on projects we don't believe will work.

Ready to start? Send a brief describing the problem and we’ll tell you plainly which step you actually need.

Get in touch →