Health coaching · productDesigned

Let customers log their day by speaking.

A health coaching product’s customers already speak their day into a separate messaging app, because typing it in is the step they drop first. The system is drawn to turn a one-tap voice note into structured entries the team can review, tagged to the right customer from the start.

Architecture

The system.

  1. 01IdentifyThe customer is signed in before they speak, so every entry is tagged to the right person from the start.
  2. 02SpeakOne tap, and the customer says their day in their own words, in their own language.
  3. 03TranscribeAn accent-ready engine turns the voice note into text, tuned for the accents a prior attempt got wrong.
  4. 04StructureThe free speech becomes named fields: readings, meals, symptoms and mood.
  5. 05ConfirmThe customer reviews what was heard before anything saves.
  6. 06DeliverClean, customer-tagged records reach the team’s own system through an API, on their own schedule.
The boundary · never invents an entryThe system structures and tags, and never edits or invents an entry; nothing saves until the customer confirms what was heard.

The situation

Customers who stay long enough to matter are already narrating their day by voice into a messaging app, and someone on the team pieces it together into records by hand, with gaps every time a name or an accent gets missed.

A previous attempt at voice logging dropped records outright when it could not parse certain names and accents, so entries never reached the right customer’s file. Typing the day into a form removes that risk, and reintroduces the exact friction that made customers stop logging in the first place.

What the design settles

This system is designed, not running, so what follows is what the design decides, not a result.

  • Which fields a spoken day turns into, and what separates a symptom from a routine note.
  • How a customer is identified before they speak, so an entry can never land on the wrong record.
  • What the customer sees to confirm before an entry saves.
  • What crosses the API to the product’s own system, and what stays inside the log.
Demo · synthetic dataOpen the demo

Something like this, for your company?

The first conversation is with the people who would build it. If there is a fit, the Consult comes next.

Map something similar