E /Economics
The economics of governed AI.
AI tooling only pays back when people know how to use it. Put your own numbers into the calculator below (rate, hours, team size, multiplier, tool cost) and read off what the maths says, net of what the tooling costs to run.
01 /The thesis
Three honest claims.
-
01
The tools are good enough to matter. Reading, drafting, summarising, scaffolding code, extracting structured data — all faster when the user understands the tool and the review bar.
-
02
The savings disappear when usage is poor. Bad setup, weak prompts, no project context, and no review habit turn useful technology into expensive slop.
-
03
Education and governance recover the value. Shared setup, reusable skills, human review, and evidence trails make the leverage repeatable.
02 /Calculator
Run the numbers against your rate.
Browser-side, no data leaves your device. Move the slider, edit the numbers, watch the year totals update. The net figure subtracts an honest AI-tool cost — the maths is the maths.
03 /Why the gate exists
One error can erase the year.
The calculator above shows the upside. This table shows the downside, and why the human approval gate is part of the economics, not a tax on them.
| Scenario | What goes wrong | Indicative cost |
|---|---|---|
| Bad VAT close | AI mis-categorises a quarter of receipts. Filing is wrong. Re-work, penalties, and trust loss with the practice. | £20k+ |
| Mis-sent comms | Drafted email goes out without the source-bound check. Names misquoted, claims uncited. Customer escalation. | £30k+ |
| Regulatory miss | Compliance-sensitive output reaches a customer without the evidence pack. Audit window expires. | £80k+ |
| Vendor-tool churn | Wrong AI subscription strategy. £15k/yr in seats no-one uses; team loses appetite for adoption. | £15k+ |
Read this against the calculator: a single high-impact error can wipe out the full year's calculated saving. The gate is cheap relative to the tail.
04 /Where the numbers come from
The honest fine print.
46 working weeks per year. Accounts for holiday, sick days, and bank holidays in a UK context. If your team works more, your number is higher.
FTE benchmark £100k loaded. Salary + employer NI + pension + overheads for a mid-senior technical role in the UK. Adjust mentally if your role costs less or more.
Multiplier range 1.5×–5×. Use the low end for early adoption and the high end for a tuned, repeatable workflow. The default 4× keeps the model useful without pretending every process behaves the same.
Not counted. Time spent learning the workflow, error-recovery cost, cost-of-error tail (see table above), model API spend beyond the subscription tier. The calculator is intentionally under-counted on both sides — the up- and the downside both compound from the visible numbers.
·Next
Run these against your real numbers.
A 30-minute discovery call is enough to ground these defaults in your team's actual rates, workflows, and constraints. That's where the spreadsheet meets reality.
We'll bring the workflow audit. You bring the rate sheet and the one process you'd most like to stop doing manually.
Book a discovery call