Voice-first onboarding: getting makers onto a system they can't read

The form is the barrier, not the technology

Programmes usually respond by hiring data-entry staff, which works until the programme scales or the funding ends. The dependency was never the software — it was the requirement that knowledge be typed.

What voice-first intake actually looks like

The maker takes photographs of the object and records a short spoken description in whatever language they think in. That is the whole submission. Everything else is derived.

• Photographs from any camera phone, no styling required
• A spoken description, any language, no script
• Optional: quantity, and how long the piece takes to make
• Field staff can submit on a maker's behalf when needed

What the system has to do with it

Transcription, translation and structured extraction turn the recording into candidate fields: title, description, material, technique, dimensions, care instructions and a price band. This is a draft, explicitly marked as one.

Extraction should be conservative. A system that guesses a material it did not hear is worse than one that leaves the field empty for a reviewer.

Designing the review queue

Reviewers need three things: the original photographs and audio alongside the draft, a single-key path to approve or send back, and a record of who decided what. Anything slower than roughly two minutes per submission will not survive contact with a real programme.

Rollout sequence that works

Start with ten makers and one reviewer, not the whole membership. Confirm the review time per submission, then confirm the first payable settles correctly, then widen. Programmes that onboard hundreds of members before the first sale end up with a catalog nobody has verified.