How we work
Our five-phase process for AI projects
We work in five phases: assess, prioritize, pilot, adopt, expand. We start from your business problem instead of a piece of technology, prove value on one focused build before anyone commits to more, and hand over systems you own outright.
The five phases
What happens in each of the five phases
Opportunity map
Assess
Understand how the business works and where the biggest opportunities are.
We talk to leadership and to the people actually doing the work, and look at the systems information moves between. The output is a picture of where the hours go, which is usually not where people assume.
Prioritized roadmap
Prioritize
Rank those opportunities by value, feasibility, effort and risk.
Every candidate gets weighed against the cost of leaving it alone. Some things that look exciting do not survive that comparison, and that is the point of doing it before anyone builds.
Tested pilot
Pilot
Start with a focused implementation tied to one real business problem.
One problem, one measurable outcome, a defined timeframe. Small enough that failure is cheap and success is obvious.
Team runbook
Adopt
Help the team understand and use the technology in their everyday work.
A system nobody opens is the same as no system. This phase is training, adjustment, and watching what people actually do rather than what the demo assumed they would do.
Next decision
Expand
Build on what works and decide together what comes next.
With one result on the board, the second decision is much better informed than the first. Some clients keep going, some take the roadmap and run it themselves.
Built for ownership
You leave with something you can run.
A system in your accounts, documentation your team can use, and a verified handoff.
Architecture & infrastructure
Security documentation
Acceptance criteria
Runbooks & credential transfer
How we differ
Six principles behind how we work
We show what it costs to do nothing
01Every recommendation is presented against the cost of leaving things exactly as they are, because that is the real alternative. If the work costs six hours a week today, that number sits next to the fix. Sometimes it makes the case and sometimes it kills it, which is the point.
We will tell you not to build
02Not every process needs automation. Some problems are better solved by changing a workflow, switching on a feature you already pay for, or dropping a tool you have outgrown. A recommendation to do nothing is a legitimate outcome of an assessment, and we would rather give it than sell a project.
You own what we build
03Systems run in your accounts, on your infrastructure, with your API keys, handed over with a live walkthrough. You are not renting access to something we control, and nothing stops working if you stop working with us.
We use what you already pay for first
04The cheapest win is a capability you already own and never switched on. That is the first place we look, before anyone talks about buying or building something new.
We are vendor and model agnostic
05We build with Claude, ChatGPT, Copilot, Azure, or a local model, based on what fits the job and your constraints. We do not resell any of them, so the recommendation is not shaped by what pays us.
We start with one problem
06One worthwhile problem is enough to start. Massive transformation programs are how AI budgets get spent without anything changing.
Common questions
Questions about how we work
Assessment timing depends on the scope agreed at kickoff. A first pilot is typically scoped to around sixty days, because that is long enough to build something real and short enough that you find out whether it worked.
You have a prioritized roadmap and no obligation. Some clients have us build the first item, some hand it to their own team, and some sit on it for a quarter. The roadmap is yours in all three cases.
For clients with internal technical capacity who want a second set of eyes rather than a build partner, yes. Most engagements start as a defined project instead, because open-ended retainers make it too easy for both sides to lose track of whether anything is working.
Then we can scope that directly. We will still ask what problem it solves and what it costs you today, because occasionally that conversation changes what gets built, and it is a cheap conversation to have first.
Find out where AI can save your business time and money.
Find your stage with the free roadmap. Then decide which opportunity is worth pursuing.
Free · No sales pitch · You keep the roadmap