Answers you can say out loud.

A rep who repeats a wrong number to a physician loses the room. Patency's assistant is built so that every number comes from the data, not the model.

Illustrative data

How grounding works

The model reads the question. The data writes the numbers.

Step 1
You ask
“Who near me is growing fastest in thrombectomy?”
Step 2
The model plans
It turns your question into a structured query. It does not write any numbers.
Step 3
The platform answers
Values come straight from Patency's data, with source, time period and measured or estimated labels.
Step 4
Checks run
If any check fails, you get a plain table of results instead of a written answer.

Built for the field first, with the controls a compliance team expects.

Ask
Plain-English questions answered as tables and cards in seconds.
First release
Runs with defaults
No interrogation. It answers first; you refine.
First release
Audit log
Every AI and data action is logged and visible to your admin.
First release
Patency inside your AI tools
Connect to the assistants your company already uses, with the same data, the same permissions and a full audit log.
Coming
Actions with approval
Ask it to build a list, plan a route or push to CRM. Nothing writes without your one-tap approval.
Coming
Admin AI policy
Per-company policy and an approved-client allowlist.
Coming

It won't present an estimate as a measured fact.
It won't change your CRM or lists without your approval.
It doesn't use patient records. Patency works with provider- and site-level aggregate data.
It doesn't give clinical advice. It's a sales planning tool.

Bring a question you'd ask on a Monday morning. We'll show you the answer, and where every number came from.