Metered per question
Who asked, which dashboard or report it belongs to, how many tokens it consumed, how long it took and what it cost.
Ask
The report nobody built in advance
Six months after any reporting project goes live, someone asks a question nobody anticipated. That is normally where these products start to go stale. Prism is built so that moment is routine.
Type the question. Keep the answer.
Someone in management types a question in plain English and gets back a table or a chart, built from your own data. If the answer is useful they pin it, and it becomes a permanent dashboard tile that runs every month from then on.
That is the part that matters commercially: your requirement list does not have to be complete on day one, and a new question does not become a change request.
What makes the answers trustworthy
- The model never sees your raw hospital schema. It sees a curated set of business views with clean names and written definitions.
- Access rules are applied by the platform after the query is generated, never requested of the model — a clerk asking for every doctor's payout still gets only what their role permits.
- Only read queries can run at all, enforced by database permissions, not by instructions.
- Senior users can see the exact query behind any answer.
- Every question, the query it produced and its result count is logged.
Illustrative. Figures shown are sample data.
Cost control
You can see exactly what the AI costs
Most vendors will not show you this. Prism meters every call and puts the log in your administrators' hands, not only ours.
A monthly allowance
Each plan carries an allowance you can watch against. If it runs out, the Ask box pauses — every dashboard keeps working.
Per user and per dashboard
See which team is using it and which reports are expensive to run, so the heavy ones can be pinned and scheduled instead of re-asked.
A question that gets asked every month should be pinned once, not retyped thirty times. The token log is what makes that visible.
Examples
Questions hospitals actually ask
Which wards ran below 60% occupancy on weekdays last quarter?
Show discount percentage by consultant for cash patients this financial year.
Which items expired in stock last month and what did they cost us?
Compare average length of stay by specialty against the same month last year.
List admissions with no final bill raised, oldest first.
Which referring doctors sent fewer patients this quarter than last?
Try it against your own data
The pilot includes the Ask box, so your team can test it on real questions before you commit.