Every department bought its own AI.
Business units control most software spend, so six to eight overlapping AI apps is normal. Bills get surprised.
Platform £4,000 to £14,000 a month · audit £12,500, credited
of UK leaders have only partial visibility of AI spend (KPMG UK, 2026)
of AI cost now goes to premium models, up from 6% in ten months (Ramp, 2026)
of IT leaders have been hit by unexpected AI charges (Zylo, 2026, vendor survey)
First findings within 48 hours of your health check.
Issues detected · needs attention first
| Issue | Rule | Severity | Owner | Status |
|---|---|---|---|---|
| Payroll file uploaded to an external AI tool | R-15 | High | HR Director | Blocked · reviewed |
| 31 personal AI accounts found in Sales | R-02 | High | Head of Sales | Migrating · 27 done |
| Customer data in a prompt without redaction | R-01 | High | Ops lead | Redacted · 0 left |
| Premium model used for internal drafts · 412 requests | R-07 | Medium | Engineering lead | Rerouted |
| CRM "AI assist" running without a DPIA | Policy | Medium | Head of Sales | DPIA in progress |
| Agent attempted an external email | R-12 | Low | S. Reid | Held · approved |
| Job | Owner | RAG | Effort | Cost to date | Forecast | Actual | Variance | Return / yr | Status |
|---|---|---|---|---|---|---|---|---|---|
| Migrate personal accounts to managed | IT Director | 12 d | £4,200 | £6,000 · 20 d | £4,200 · 12 d | −£1,800 · −8 d | risk | Done | |
| Route everyday tasks to standard tier | FD | 3 d | £900 | £1,200 · 4 d | £900 · 3 d | −£300 · −1 d | £58,400 | Done | |
| Payroll upload block (R-15) | HR Director | 1 d | £300 | £300 · 1 d | £300 · 1 d | 0 | risk | Done | |
| Collections letters via Comms agent | Head of Collections | 18 d | £11,600 | £14,000 · 25 d | — | on track | £96,000 | In progress | |
| KYC document checks | Head of Onboarding | 9 d | £5,100 | £9,000 · 15 d | — | on track | £61,000 | In progress | |
| Contact-centre adviser assist | Head of CS | 26 d | £19,800 | £16,000 · 20 d | — | +£3,800 · +6 d | £142,000 | At risk · vendor delay |
Effort, cost and return are recorded against each job as it happens. Done means the job is solved in production with its owner’s sign-off, not that a licence was bought.
We test it before you use it.
New models arrive most weeks from a handful of companies. Deciding which to use, for what, and when not to, is not a decision your employees should be making. So we make it first, in a controlled sandbox, with a dedicated research team.
Ask us to test a specific frontier model against your workload. You get a written result and, if it passes, a staged switch with a way back.
| Release | Behaviour | Stability | Cost vs current | Data | Verdict |
|---|---|---|---|---|---|
| Standard-tier model, 9 Sep | Pass | Pass · 0.4% drift | −11% | Pass | Approved |
| Premium model, 2 Sep | Pass | Fail · 6% drift | +38% | Pass | Declined |
| Small model, 11 Sep | Testing | Day 3 of 7 | −54% | Pass | In progress |
| Client request · voice model | Testing | Day 1 of 7 | — | Review | In progress |
Example data. Results are written up monthly in the Change Control briefing.
From businesses your size.
Our IT team is stretched. Who does the work?
We do, with your sponsor. Deployment is into your cloud account, by our engineers, with your change process. A customer success manager runs the audit and the 90-day plan.
We already use Copilot across the company.
Good. Copilot stays. The gateway puts your rules, redaction and metering in front of it and everything else, so Copilot use appears in the same register and the same bill.
How do we justify this to the board?
With the value register: every job solved, what it cost, what it returned, and a variance column. The audit gives the first version inside ten working days.
Find out where your AI stands. Then decide.
Under a minute, an upload, or a visit. Something for the board either way.