AI Clinical History

The Problem

The main problem I was trying to solve here has to do with the clinical history of a patient. These records often present some if not all of these issues:

• Large amount of medical data. A patient's clinical history can be quite long, going back years, finding specific information can mean sorting through records.
• Inconsistent data entry. Most clinics do not have a template to follow, meaning each vet writes clinical data differently.
• Doctors need to manually search for relevant information. With so much data available, it is difficult to pinpoint relevant information needed for that day's visit.

What I Did

Using Provet's API and Make.com I created an integration that allows a user to summarize the history of a patient. The logic is pretty simple:

• When a user goes to the patient history panel, they can select the "Patient History Summary" button.
• On click, a webhook will send a request to Make.com with the patient's data and instructions for ChatGPT.
• ChatGPT will analyze the data, and return the summary, following a structure I gave as instructions.
• Finally Make.com will send back to the API the content, which is loaded in a drawer the user can see.

AI prototyping

The amazing Guilherme Dumas — fellow product designer at Provet, had the fantastic idea of using Cursor AI and our front-end repo to create fully interactive prototypes.

Every designer knows the limitations that prototyping tools like Figma carry. It is almost impossible to mimic complex interactions (like editable inputs, scrollable tables or filtering toolbars). It also makes complex flows result in a mess of linking arrows. Overall, these tools simply do not scale well.