Skip to content
SMZone

A system, a runbook, and the keys.

There is no SMZone platform to licence and no seat to buy. We build a decision system inside your infrastructure, run it with you until it is boring, then hand it over. What follows is the whole deliverable.

A control desk of analogue meters and switches in a generating station

01The deliverable

Five things

What lands at the end.

Five things, every time. If an engagement cannot produce all five, we say so in week one.

  1. The running system

    Deployed in your cloud account, behind your own auth, reading your own data. Not a hosted endpoint we control.

  2. The integration

    Output wired into the interface where the decision is actually made — the CRM, the dispatch board, the ledger, the queue.

  3. Monitoring

    Input drift, decision volume, latency and model version, on a panel your team can read without us.

  4. The runbook

    What to do when it breaks, who to call, how to roll back, and the conditions under which it should be retired.

  5. The code

    Every line, in your repository, with the training pipeline and the tests. No black box, no escrow clause.

02Integration

It goes behind what you already use.

Racking aisles running the length of a distribution centre

Nobody wants another tab. The systems we build surface inside the tool your team already has open — a score in a column, a sorted queue, a suggested route, a flagged line in the ledger. If a person has to change how they work to use it, we have built the wrong thing.

Where it reads from
Your warehouse, your operational database, your event stream, your document store. We work with what exists rather than asking for a migration first.
Where it writes to
The system of record you already operate. Every decision carries its model version and its inputs, so any output can be explained months later.
What it does when unsure
Escalates. Confidence thresholds are set with your team, and below them the system hands the decision to a person rather than guessing.

03Operating

Handover

Who runs it afterwards.

Your team. That is the point. We stay on for a defined support window, then step back to an annual review. If you would rather we kept operating it, that is a separate conversation and a separate contract — not a dependency we design in.

  • Taped

    Running in production, inside its agreed error budget.

  • Struck

    Drifting or failing its threshold. Pulled back to the bench for rework.

  • On a pin, deferred

    On a pin

    Built and tested, not yet committed. Reachable, costing nothing.

Tell us about one decision.

We will tell you within two weeks whether your data can support it.

Start a pilot