How we work

Context first. Then capability. Then autonomy.

There is a fixed order to this work and we do not skip steps to make a sales cycle shorter. An agent with no context is not useful, it is just fast at getting something wrong. So we build the context first, add capability on top of it, and only then hand over autonomy — never the reverse.

The order, and what each step holds

1 · ContextConnectors, identity, permissions and entity resolution. A connected layer is 3–5 months.
2 · CapabilityA governed assistant and automations, built as each source comes online. First automation in 2–4 weeks.
3 · AutonomyAgents start at L2 Propose. A task reaches L3 only on its own shadow-mode record.
FloorL4 always: money, safety and people never act without a named person.
Measured onHours, quote turnaround, rework, cash days, on-time delivery. Baselined first.

Planning ranges, not a quote. The state of your data moves them more than the choice of model does.

Autonomy tiers: L1 Assist · L2 Propose (the default) · L3 Act within written bounds · L4 Approval required always. What each tier means

1The arc

An honest maturity model.

We use four stages to tell you honestly where you sit today, not where a demo makes you feel you sit. A licence and a few automations is Stage 1, not Stage 2. Nobody reaches Stage 2 without deliberately building the context layer.

Fig. 1The four stages, and the test for each

Stage 0
AI Zero
Nothing sanctioned. Staff are using consumer AI anyway, and nobody can see what went in.
The test
You cannot answer what was pasted into a chatbot last week.
Stage 1
AI Enabled
Licences bought, a governed assistant in place, a handful of automations running. Useful, but the model still starts cold on every task.
The test
AI saves minutes, not decisions.
Stage 2
Context Connected
The context layer is live. Systems are joined, permissions resolve per request, and answers cite where they came from.
The test
You can ask your business a question and get an answer with a source.
Stage 3
AI Native
Agents hold standing jobs on top of the context layer. Each task starts at L2 Propose, moves to L3 within written bounds only on its own shadow-mode record, and money, safety and people stay at L4. The record is complete.
The test
Work finishes while nobody is watching, and you can audit exactly how.

Autonomy tiers: L1 Assist · L2 Propose (the default) · L3 Act within written bounds · L4 Approval required always. What each tier means

2Honest timelines

Real shapes, not sales timelines.

2–4 wks

First working automation, in build increment one

4–8 wks

Governed assistant + automations live

3–5 mths

Connected context layer

6–12 mths+

Agents carrying standing work, staged

These are planning ranges for this kind of work, not a guarantee we will quote you in a proposal. The variable that moves them most is not which model you use. It is the state of your data — a business with clean systems and a named data owner moves through these stages faster than one that has never audited what a system actually holds.

3How we measure

We baseline before we build.

Not because it looks rigorous. It is the only way either of us will know whether a stage worked.

If a stage will not pay back, we say so before we build it, not after you have paid for it. That is the point of measuring first, and we would rather cut a stage than pad a roadmap with one that will not earn its cost.

01
Hours by function
Where time actually goes today, by team, before anything is automated or assisted.
02
Quote turnaround
How long it takes from request to a number the customer can act on.
03
Rework rate
How often a job, quote or record has to be redone because it was wrong the first time.
04
Cash days
How long cash takes to move from work done to money in the account.
05
On-time delivery
Whether what was promised arrived when it was promised, without the caveats.

4What you own

Your cloud. Your accounts. Your data. Your code.

No lock-in, and no proprietary runtime you cannot leave. If you replace us, what we built keeps running under your own operations team.

Your cloud

Infrastructure runs in an environment you control, not a platform we operate on your behalf.

Your accounts

Every connector authenticates as a resource you own and can revoke, on your terms, at any time.

Your data

Nothing trains a model outside your tenancy without a separate, explicit agreement to do so.

Your code

The context layer and everything built on top of it ships to a repository you own and can read.

5What we need from you

This works better if you bring four things.

None of them are a budget line. All of them are harder to fake than a budget line.

01
An executive sponsor who can make decisions
Not someone who attends the meetings — someone who can say what gets built, what gets shelved, and when a process is the problem rather than the tooling.
02
Access to the systems, early
Waiting until week three to get us into the ERP costs you a week of an engagement you already paid for.
03
About two hours a week from someone who knows how the work really happens
Not the manager who describes the process — the person who does it, exceptions and workarounds included. That is roughly 6 hours across an AI Review and 10 across a Strategy. A build phase names its own figure before it starts.
04
A willingness to hear that a process is the problem, not the tooling
Sometimes the honest finding is that a workflow needs to change before software can help it. That is a harder conversation than buying a licence, and we will still have it.

See the three engagements, or just start talking.

The order above is how we operate once we are working together. Where you enter it depends on what you already know about your own business.

See the three engagements Book a fit call