Tell us how your AI actually works.
A short questionnaire we send before a purchase decision. It is deliberately non-technical: we want to understand what is under the hood and how it is priced, not audit your architecture. Wherever you are not certain, say so — an honest “not sure” is far more useful to us than a confident guess.
01 Your product
The basics, and how widely it is already running.
What does it do for a customer like us?
One or two plain sentences. No architecture needed.
How many customers are live on this in production today?
Live and in daily use, not pilots or signed-but-not-deployed.
Will it touch patient or personal data?
At any point in the workflow, including logs.
02 What kind of AI is inside
These two answers decide which of the remaining sections apply to you.
Does it use a general-purpose model from a major AI lab?
The kind you license rather than build.
Does it use models you trained yourselves?
Including anything tuned or fine-tuned on customer data.
03 The general-purpose models
Where they run, whose they are, what each one is for, and what the tokens cost.
Which cloud does it run in?
Labs change far less often than model names, so this is the question that stays true. Tick all that apply.
Which models do which jobs?
Describe the jobs rather than the version numbers — those change every few months. We want to know where the heavy reasoning happens and where you use something small and cheap.
How is model usage charged to us?
The token question. We are trying to understand whether inference cost sits with you or with us.
Is there a ceiling on that cost?
Can we see our own token usage?
What does a typical unit of work cost in tokens or dollars?
However you measure it — per encounter, per case, per document, per user per month. A rough figure and the unit is enough.
If your model provider changes prices, who absorbs it?
Is customer data used to train or improve the models?
Where is the data processed?
04 The models you trained
Where the data comes from, how often it is refreshed, and who does the work.
Tick all that apply.
How frequently do you train?
Who builds and validates these models?
Is one customer’s data kept separate from another’s?
Anything else about the data
Whose it is and under what agreement. Optional.
05 Control and paperwork
What it does on its own, and what is already signed.