TENURED BY IGNITE9 · AI COST AND MODELS
Pay for the thinking, not for the routine.
Tenured sends a model only the work that needs one. Everything else runs as rules, reviewed routines, code and API reads, and every model call that does happen is on a ledger you can check.
01 / WHY THE BILL STAYS SMALL
Three levers, set by your organization.
Deterministic first
A rule, a reviewed routine, a piece of code or an API read comes before a model call. Work that needs no model never reaches one, and repeated work becomes a routine your administrators review.
Machine routes before screens
A connected vendor MCP server or API comes before the application’s screens, so a digital employee reads what it needs in one call instead of many steps.
The right model for each role
Your organization chooses the model each digital employee uses, with backup models that take over when a provider fails.
02 / WHAT YOU CAN CHECK
Every model call on the record.
The figures come from your organization’s own records, so your finance team can check each one.
A cost ledger of every model call
Each model turn and each verification writes one row: who caused it, where it went, the tokens the provider reported and the model that answered, priced at the price in effect then. Rows are never changed afterwards, and they hold ids, counts and money only, never content.
Your own contract prices
Your administrators record the prices in your contracts with model providers; they apply to your organization only, and under four-eyes a second administrator approves them.
An efficiency report
The work Tenured did without a model call, by source: rules, reviewed routines, code, API reads, waits handled below the model and vendor channels. Every estimate names its basis, taken from your organization’s own measured model work.
Screens, counted
The report shows the share of missions that opened an application’s screens, overall and per application, next to the measured model work of the missions that used a machine route instead.
Budgets with alerts
A monthly model budget per organization, with alerts at 50, 80 and 95 percent of it.
Your keys, your bill
Your organization brings its own model keys, so the model provider bills you directly, and the ledger shows each call against it.
03 / YOUR CHOICE OF MODELS
The model is your decision.
Choose Claude, OpenAI, Gemini, DeepSeek or another model with an OpenAI-compatible API, including one you host yourself, once its address is on the service’s approved list. Each digital employee can use a different one, and backup models you register keep the desk answering when a provider is down; only backups inside your data boundary qualify.
Models and your data boundaryTHE MODELS YOU CAN USE
- Claude
- OpenAI
- Gemini
- DeepSeek
- Any model with an OpenAI-compatible API
- A model you host yourself
04 / ESTIMATE YOUR RETURN
Put your own numbers in.
The estimate uses only what you enter and shows its formula. In an engagement, Tenured’s own value overview and efficiency report replace your assumptions with your records.
- Time returned to your teams
- Enter the first four values
- Value of that time
- —
- AI model spend avoided
- Optional
Hours returned a month = hours × agent share × (1 − review share)
Value a year = hours returned × cost per hour × 12
Model spend avoided a year = model spend × repeated-work share × share moved × 12
An estimate from the values you entered, not a quote or a measured result. It leaves out license, implementation and change costs, and your results depend on your work, systems and approvals. We will gladly build a real business case with you in a demo.
Tenured’s capabilities are delivered through design-partner engagements: scope and timing are confirmed when we scope your work.
NEXT STEP
Bring one ticket that bounced between five teams.
In a 30-minute demo we show Tenured at work, walk through how it would handle yours, answer your security team’s first questions and agree the scope of a design-partner engagement.