Routine replies drafted in seconds.Anything clinical goes straight to a person.
A working concept of a patient-support desk, built on fictional patients, orders and messages, so everything is safe to click.
Built by Chanon Poovaviranon (Beam) as a concept for Dispensed. Not affiliated with Dispensed.
Two sample messages, run through the checks
Loading the live example
Two messages from the demo queue, through the same checks the desk uses.
What keeps it safe
Safety rules run before any AI.
Every number in a draft is checked.
Dates, amounts, order numbers and tracking numbers must match the patient's own records. If one does not, the draft is blocked and a person writes the reply.A person always sends.
The AI only drafts routine replies. An agent reads each one, then sends, edits or escalates it. Nothing goes to a patient on its own.
Built for your team
This demo runs on fictional data. Running it on your own helpdesk would look like this.
- Weeks 1 to 2Connect your helpdesk, order records and support policy, and agree the safety rules with your clinical team.
- Weeks 3 to 4Order-status drafts go live for one agent. A person sends every reply, and the safety test set runs on every change.
- Days 31 to 90More routine message types join, the weekly numbers are compared with the figures from before, and the same approach is scoped for clinic and pharmacy work.
Talk to Chanon Poovaviranon (Beam)
He designed and built this concept, from the safety rules to the test set, and would set it up with your team.
Or write to chanonbcp@gmail.com