Onhold

Onhold is a concept AI assistant that calls the gym, the bank or the dentist for you, waits through the hold music, talks to a person and texts you when it's done. I wrote the positioning and every line of copy, built the identity, designed and coded the landing page with a live call demo, and designed the texts and emails that follow. Onhold is not a real product.

The buyer is anyone who has put off a phone call for weeks. Signing up for a gym takes one tap; cancelling takes a phone menu, working hours and a long hold.

The page had two jobs: make the product feel light and a bit funny, and answer the obvious worry, "what will it say on my behalf?" Success meant a visitor could see, without reading, that Onhold waits, talks and reports back.

Hold music, made visible.

Voice AI products usually look like dark dashboards or purple gradients. This one is a consumer app about a shared annoyance, so the identity is warm and a little silly: a looping waveform, round shapes, and the colours of a 1970s rotary phone on white. Gabarito sets the headlines.

The hero plays a whole call in about twelve seconds. It presses 3 at the menu, the hold timer races to 47 minutes, the agent offers a discount, Onhold says no, and a text arrives. The product is explained before anyone scrolls.

You only hear from Onhold when it needs you or when the job is done: one call, four screens of texts.

Three colours, three voices. Butter is Onhold talking, ink is you, and tomato appears only when a decision costs money, which keeps the call on hold until you answer.

The waitlist email asks for a list of calls instead of a click, the call summary is a receipt you can reply to, and the recap shows the hold time you got back.

The first motion cover reused the move I'd made for Meterline: a logo chip grows into a window, tilts in 3D and lands on a phone. Side by side with the other covers it read as a template, so I dropped it and made Onhold vertical and mobile-first.

Real callers, real call centres and a waitlist. If it shipped, I'd measure how many visitors watch the demo to the final text before joining, and whether reading the rules section changes how many people join.