Our Process
- Home
- Our Process
How we work
Two weeks to clarity, sixteen to production
Every engagement runs the same three phases. The first one can end with us telling you not to build the thing — that is the point of running it first.
01
Two weeks
Readiness Review
Before anything is built, we work out whether it should be. Two weeks mapping your data, your decisions and your risk appetite.
What happens
- Map where the data actually lives and what state it is in
- Trace who makes the decision the system is meant to support
- Test each candidate use case against effort, value and risk
- Agree what "working" will mean, and how it gets measured
What you are left with
- A scored shortlist of use cases, with the weak ones named
- A sequenced roadmap with dependencies made explicit
- A written view of the data and governance work required first
02
Eight to sixteen weeks
Build & Integrate
The model, the pipeline and the guardrails, deployed inside your environment and wired into the systems you already run.
What happens
- Build and evaluate inside your own tenancy, not ours
- Wire into the systems your teams already use day to day
- Design the review path for anything the system is unsure about
- Instrument it so drift and failure are visible, not discovered later
What you are left with
- A working system in production, integrated rather than bolted on
- An audit trail covering inputs, outputs and human overrides
- A model card: what it does, where it fails, who reviewed it
03
Runs alongside delivery
Enable & Hand Over
Role-based training runs in parallel with the build. By go-live your own people hold the runbook, the dashboard and the on-call rota.
What happens
- Train by role, so each team learns the part they actually own
- Name an internal owner early and build the handover around them
- Write the runbooks with your team rather than handing them over
- Rehearse the failure cases before they happen in production
What you are left with
- A trained internal owner accountable for the system
- Runbooks, monitoring and an on-call rota your team runs
- No standing dependency on us — by design
What holds across all of it
Four things we do not trade away
We will tell you not to build it
If a problem does not need AI, the readiness review says so. That is a cheaper answer to receive in week two than in month nine.
Your environment, your data
Systems are built and run inside your own tenancy. Residency and governance constraints are inputs to the design, not exceptions handled later.
Fixed scope, fixed date
Every engagement has a named team, an agreed scope and a date. An engagement should not get more expensive the longer it runs.
Built to be handed over
The measure of success is your team running it without us. Training, runbooks and ownership are part of delivery, not a closing formality.
Start with the readiness review
Forty-five minutes, no charge, and an honest answer on whether this is a problem AI should be solving.
Book a Consultation