Service · AI Ops, Governance & Gateway

Know what AI your company is using, and what it touched

Most companies discover their AI usage from a credit card statement. A gateway puts every model call through one door with a key, a budget, and a log, without blocking the people doing useful work.

  • One gateway, one key per team, one bill you can read
  • Hard spend caps rather than after-the-fact surprise
  • Every call logged: who, what model, what it cost
  • A written policy short enough that people read it

What you get

Deliverables, not slideware

  • 01

    Usage audit

    Which tools and models are in use today, by whom, and what they cost per month.

  • 02

    Gateway in place

    A single access point with keys per team, model allow-lists, and hard budget limits.

  • 03

    Audit log

    A queryable record of calls and what data category they involved, for security reviews.

  • 04

    Usage policy

    One page: what may be sent to a model, what may not, and who to ask.

ai-usage-policy.md
  • 01allowed: internal docs, anonymised customer records
  • 02never: full payment data, unredacted contracts
  • 03budgets: support $200/mo · ops $350/mo · hard cap
  • 04logging: every call retained 90 days
  • 05exception route: named owner, 1 business day
Sample of the deliverable

How it runs

What actually happens, step by step

No discovery theatre. Each step ends with something you can read or use.

  1. 01

    Find the shadow usage

    Subscriptions, keys, and browser tools already in play.

  2. 02

    Route through one door

    Keys per team, allow-listed models, budget caps.

  3. 03

    Log and attribute

    Cost and usage per team, per workflow, per month.

  4. 04

    Publish the policy

    One page people actually follow.

Tangible

What lands in your hands

Named objects with a format and a week attached, so the handover is easy to picture.

  • Working automation

    One process running end to end, on your data, with a human approval step.

    Deployed servicePhase one
  • Operating runbook

    What to do when it fails, who to call, how to roll back.

    PDF · 8 pagesFinal week
  • Evidence dashboard

    The numbers that prove it worked, refreshed daily.

    Live URLWeek 5
  • Source repository

    Yours from day one, including the prompts and the evaluation set.

    Git repo + READMEWeek 1
Sample

Operating Runbook

Failure modes, rollback steps, and who owns each alert.

8 pp
Download the sample (PDF)

This is the real template, with client figures replaced by representative ones. Read it before you talk to us — if the format is not useful to you, the engagement will not be either.

Before and after

What the change looks like in the week

Red is the cost you carry today. Green is what the system gives back.

Today

  • Personal accounts and keys nobody tracks
  • A monthly AI bill with no owner
  • No answer when a customer asks what data models see
  • Policy that exists only in a Slack thread

After

  • One gateway, keys issued and revocable
  • Spend capped and attributed per team
  • An audit log you can hand to a reviewer
  • A one-page policy people can follow

Objections

Answered before you ask

Will this slow our team down?

It should not. The gateway is a drop-in endpoint. The friction people notice is the budget cap, which is the point.

Do we have to standardise on one model provider?

No. The gateway sits in front of several, which is usually cheaper because you can route cheap work to cheap models.

Is this compliance certification?

No. It gives you the evidence a reviewer asks for. Certification programmes are outside what we do.

One next step, and it is a paid one on purpose

The AI Opportunity Diagnostic is a fixed-scope engagement. You leave with a ranked plan you can act on with us or without us.

Still sizing the problem? Write to us in your own words first. A person reads it.