Tokenokio for consultancies
Know what AI costs per client — before it eats the engagement margin
AI is now a real delivery input, and on fixed-fee work it comes straight out of margin. Most firms cannot say what a given engagement consumed, so it is never recovered and never priced in. Tokenokio attributes usage to the project that spent it and lowers the unit cost of delivery.
What this fixes
- AI delivery cost disappears into overhead
- Usage is attributed per project, which for most firms is per client — exported as a per-engagement CSV you can attach to a bill or a margin review.
- Can't price AI into a proposal with any confidence
- Price the same task across every model in the catalogue before you commit. The spread between the cheapest capable model and the frontier default is routinely large enough to change a fixed fee.
- Unprovable savings claims to clients
- Proof of Savings separates measured, banked dollars from estimated ongoing savings and never mixes them — so what you show a client survives scrutiny.
- Juniors defaulting to the most expensive model
- Routing makes the cheap-but-capable choice the default and escalates automatically when quality is weak, so cost discipline does not depend on individual judgement.
- Client data sensitivity blocks new tooling
- Nothing is uploaded. Client prompts and outputs never reach us, which is usually the shortest answer to a client's security questionnaire about your tooling.
Before you ask procurement
Client confidentiality
Usage data is parsed in your browser and stored on your device. We never receive your prompts, outputs, API keys or payment details, and there is no user database to breach. Independent certifications (SOC 2 Type II, ISO/IEC 27001) are on the roadmap and not yet held; what exists today is an adversarial security review with findings remediated and security behaviour enforced by automated tests on every change.
Per-engagement attribution
Project, team and machine attribution with a finance-ready export, so AI becomes a line item you can recover or price in rather than an untraceable cost.
Honest savings
Measured savings and projections are reported separately, capped and conservative. Nothing is presented as banked until it actually is.
Questions we get asked
Can we bill AI usage back to a client?
You can produce the evidence for it. Usage is attributed per project with tokens, model, date and cost, exported as a CSV showing whether each figure is provider-reported or estimated at list price. Whether it is recoverable depends on your engagement terms.
Does our client content go anywhere?
No. Scans parse in the browser and stay on the consultant's device; prompts, outputs and files are never transmitted to us, and we hold no copy. API keys are used once in memory for a call you initiate and are never stored or logged.
How much can routing actually save on delivery?
It depends on your task mix, which is why the tool measures your own before claiming anything. The structural saving comes from not paying frontier prices for classification, extraction and summarisation, and from running heavy work on subscriptions you already pay for at no extra marginal cost.
Does it work across a team of consultants on different machines?
Yes. Each consultant scans locally and exports; the files merge into one ledger with per-machine and per-project attribution. A provider admin key covers that account centrally across every device.
See one engagement's real AI cost in 60 seconds — free, local, no account.