Results Follow-up Agent · Use case
How can a hospital use an AI agent to find results nobody has acknowledged, on its own infrastructure?
A results follow-up agent reads the results and reports a hospital has issued and sets them against what happened next. It lists the results that have been reported and not acknowledged, and the follow-up a report recommends that has no order or appointment against it, each with the report, the date and the responsible clinician. The clinician decides what each one needs.
A result is reported.
Nobody acts on it.
- ✗ 01A report recommends a repeat scan in three months. Unless someone books it that day, nothing in the record will ask again.
- ✗ 02Checking every report against every order and appointment by hand is work no team has time for, so the gaps are found when a patient comes back.
Where it stops for a person
Always. The responsible clinician decides what each result needs.Always. The list is a reason to look. The responsible clinician decides, and the agent does not rank results by how serious they are.
It lists what is open.
The clinician decides.
Step · 01
On a schedule you set, the agent reads the results and reports issued since its last run.
Step · 02
It finds the ones with no acknowledgement recorded.
Step · 03
It reads each report for a recommended follow-up, and looks for an order or an appointment that matches.
Step · 04
It writes one list for each clinician: the report, the date, the sentence that recommends the follow-up, and what it could not find.
Step · 05
The clinician reads the list and decides what each one needs.
Reads from · connectors
Your own APIs
Database
Results, reports, orders and appointments, through your record system's own API or a reporting database with read-only access. Your rules for who is responsible for a result are read as a knowledge source.
Acts through · MCP servers
Nothing. It orders nothing, books nothing and contacts no patient.
What you set up
- Connect the results and appointments with read-only access.
- Write down who is responsible for a result when the requester has moved on.
- Test against a month whose outcomes you know.
Where it runs
On your own infrastructure, with a model you serve. Results and reports stay there.
How you get it today
This page is labelled Use case: the agent is built in Katonic Studio. You describe the job, connect what is named above, add test questions and release the agent once it passes them.
Does the agent decide which results are urgent?
No. It lists what has no acknowledgement and what has no follow-up booked, in date order. Judging urgency is the clinician's work.
Does it contact the patient?
No. It sends nothing to a patient and books no appointment. The list goes to the clinician you name.
What if the follow-up was arranged somewhere the agent cannot read?
It says it found no matching order or appointment in the systems it can read, and names them. The clinician marks the item as dealt with.