Prepared for Thornbury Fixed Income (demo) · 2026-08-28
AI cost — executive summary
Thornbury Fixed Income (demo) · derived from 124,800 logged calls over 24 days. Cost basis calibrated to the monthly provider spend you declared — every € below inherits that scaling. Calibration sets the cost basis; it is not a verification of any individual saving.
Annualized from the measured window.
Annualized from the measured window, on the calibrated cost basis — the delta itself is telemetry-measured.
8 recommendations Bayeto is prepared to stand behind.
7 moves held — 5 on surfaces you declared critical, 2 on surfaces whose stakes you have not declared — declaring them is what lets Bayeto judge the risk rather than withhold. Not offered — not a matter of price. Not counted in the totals above: each is mutually exclusive with a move Bayeto did recommend, so counting both would bill the same spend twice. Declaring the surfaces would release €18,883/yr of this — the figure the roadmap’s Held tile shows.
Bridge: €355,140/yr is where the run-rate stood before the changes already measured; removing €244,361/yr of measured impact leaves €110,779/yr today. The two ends are the tiles above; the arithmetic between them is this line.
What this means
Across 17 generated recommendations, Bayeto identifies €47,104/yr in direct provider-bill savings (€56,350/yr including monetized latency gains) after removing mutually exclusive alternatives. Of that, €47,104/yr is available to pursue today. A further €34,566/yr was found but deliberately withheld — 5 on surfaces you declared critical, 2 on surfaces whose stakes you have not declared — declaring them is what lets Bayeto judge the risk rather than withhold. It is not added to the figures above — each withheld move is mutually exclusive with one already counted there.
The optimization score is 75/100. Every recommendation is produced by a deterministic decision engine — the same telemetry and the same objective always yield the same recommendation. No language model participates in the decision.
How the numbers reconcile
- Gross candidates (17 recommendations)
- €129,562/yr
- − Superseded by a better move (every mutually exclusive route to the same spend — never double-counted)
- −€82,458/yr
- = De-duplicated identified opportunity
- €47,104/yr
- − Held for safety (7 moves — 5 on surfaces you declared critical, 2 on surfaces whose stakes you have not declared — declaring them is what lets Bayeto judge the risk rather than withhold — €34,566/yr, already removed on the line above as mutually exclusive)
- €0/yr
- = Actionable today (8 moves)
- €47,104/yr
- Measured run-rate impact (changes already made — telemetry-measured on the calibrated basis)
- €244,361/yr
Every line derives from the same canonical ledger as every page in the product — the briefing, recommendations, roadmap, score and graph quote these exact figures.
Track record on your own data
Before the switch Bayeto predicted €0.203/call from moving gpt-4-turbo→claude-sonnet; the measured saving is €0.221/call — 92% accurate. Compared per call, so the figure reflects the engine's prediction quality and not traffic growth (volume -2% vs the prediction's basis).
Pilot success plan
14-day windowFinish the gpt-4-turbo → claude-sonnet migration: Meeting Summaries missed the rollout
- Success criterion
- Realize at least 90% of the predicted bill saving over 14 days, with no significant latency or quality regression. That is at least €36,846/yr realized.
- How it is verified
- Bayeto reprices your actual post-change tokens at the old configuration and compares. The result is measured from your telemetry, not self-reported.
- Owner
- Next review
The 14-day window and the 90% threshold are Bayeto programme policy, not figures derived from your telemetry.
ledger version 22c35e346df765c3 — identical on every page, export, and decision record of this workspace state · how to verify it