AI evaluation and rollout
AI Cost Controls Setup
For a team whose AI provider bill has grown without a breakdown. Every key and project inventoried, spending limits set at the provider and for each feature, the cost of each request traced to the feature and customer that caused it, and the figures reconciled to the provider invoice.
- Price
- $4,900 fixed
- Duration
- Two weeks
Who it is for
- Teams whose AI provider bill has grown without a breakdown by feature or customer, typically at $5,000 or more a month in API spend.
- Finance leads who hold the AI budget, and engineering leads who hold the keys, and need one set of figures between them.
- Regulated teams more concerned with runaway or abuse risk than with the size of the bill.
What you receive
- Inventory of every API key, project and account, with its owner
- Provider limits and a gateway budget per feature, each tested to stop a request
- Cost per feature and per customer, reconciled to the last provider invoice
- Savings options ranked, with their effect estimated under stated assumptions
- One-page runbook, and a variance report 30 days after handover
Savings are estimated under stated assumptions, not guaranteed. A cheaper model is recommended only where a test set shows its answers hold. Credited in full against a Build that starts within 60 days.
Two weeks, in steps
Week 1: inventory and baseline
Every key, project, workspace and card-paid account found, with its owner, coding-assistant seats included. Three months of usage rebuilt from the provider usage data, cost per request, feature and customer where tags allow, reconciled to the last closed invoice.
Week 2: controls and runbook
Provider caps and alerts set, a gateway configured with tags for feature, customer and environment, budgets per feature with alerts, and a decision for each on whether it blocks, falls back or alerts. Each limit tested to stop a request; unused and shared keys removed or rotated; the runbook written.
Day 30: variance report
Actual spend against the budgets set, reconciled to the provider invoice, with any limit that fired or should have.
What we need from you
- Admin or read access to the provider consoles, for example an Anthropic Admin key or OpenAI organization owner access.
- Three months of provider invoices, and read access to the code that calls the models.
- Permission to set up a gateway in your cloud account, and a named owner in engineering and in finance.
Scope, price and dates are confirmed in writing before the work starts. You own what is delivered.
About the AI Cost Controls Setup
Our spend is small. Is this worth it?
Below about $5,000 a month in API spend, usually not. The provider console caps and a free gateway, set up by your own team, are enough, and the first call says so.
How much will we save?
There is no savings guarantee. Options such as prompt caching, batch requests and model routing are ranked with their effect estimated under stated assumptions. A cheaper model is recommended only where a test set shows its answers hold, which is the AI System Evaluation.
Do you read our prompts or data?
No. The cost logs keep metadata only: tokens, model, feature, customer tag and cost. No prompt content and no patient data.
What about AI inside the software we buy?
It is listed with its pricing terms, but not instrumented. The package covers API spend you control, across up to two providers and ten keys or projects.
Does DF1 run these controls itself?
Yes. The public Chart Explorer demo runs under a $5 daily model-cost ceiling. Each request's cost is recorded, and a worst-case cost is reserved before every model call, so a burst of parallel requests cannot all pass the limit.
Other packages
AI System Evaluation
From $7,500, two to four weeks. For a team with a generative AI tool in pilot or production and no measured answer to how often it is wrong.
OpenAI Team Rollout (Claude or ChatGPT)
From $4,900, about two weeks. A working setup for a team of 5 to 50: permissions checked, up to five tools connected, one workflow built for the work the team does most, and a training session on the documents the team works with.
OpenReconciliation Automation Pilot
$9,500 fixed, three weeks. One reconciliation, automated end to end on your own exports: bank to ledger, subledger to general ledger, or intercompany.
Open$4,900 fixed, two weeks. Scope confirmed in writing first.
Tell us your situation and when you need the result. A reply comes within one business day, from the person who would lead the work.