Referral orders read by AI

A patient photographs the paper referral. The AI reads it — handwriting included — and proposes the specialty, the doctor and a slot. Staff approve or reject in one click.

Status: Live at the Pinamar installation.

What the AI pulls off the paper

• The specialty or study requested, matched against the catalogue of what this installation actually offers.
• The diagnosis and the practices written on the order.
• The issue date, which is what makes an order expired.
• Whether it is signed and stamped.

What staff see

Every request arrives with a plain-language reading — looks right or doubtful — plus the fields the AI extracted and the photo it read them from. Staff never have to open a separate viewer to check a value.

It is one reading, not a jury: a single pass of one model. Its instructions are editable from the panel, and the whole queue can be worked in bulk when the readings agree.

A rejection carries the next step

The patient never gets a bare error. An expired order comes back with what to do about it; a wrong specialty comes back with a link to the right one; an unreadable photo comes back with a link to upload another. The rejection reasons are a curated list, so the answer is the same wherever it comes from.

Next: The prepared clinician