One input, every channel: a content transformation API in eight weeks

Every content team does the same unpaid work. Something is written once, then rewritten for LinkedIn, cut for X, reshaped into a carousel, restructured for a newsletter, scripted for a reel. The thinking happened once. The labour happens every time, and it is the part nobody wants to do.

The naive version of this product is a prompt with a channel name in it. That produces output that is recognisably generic, which is worse than nothing for a brand.

What we did

A fixed project: defined scope, fixed price, fixed date, delivery responsibility on us.

1. Scoping, three days. The central decision was to make a channel data rather than code. A channel is a specification of format, length, tone, structure and rendering, so adding one is a row, not a release. That is what makes a launch with more than a hundred and twenty-five of them possible at all.
2. Factory setup, week one. Environments, CI, the delivery platform, and an evaluation set of real content with human-judged good and bad outputs to score against.
3. Build, weeks two to six. Three certified engineers, shipping weekly. Agents generated and populated the channel specifications, which is high-volume and highly repetitive. The engineers owned the transformation pipeline, the rendering path for carousels, PDFs and video, and the credit accounting.
4. Hardening and launch, weeks seven to eight. Rate limiting, webhook delivery, seven-day retention, and per-element billing that does not charge for incomplete results.

What made the difference

The feedback loop is the product. Accepts, edits and rejects feed back into voice matching, so output quality climbs over the first weeks of a customer's use rather than being fixed at signup. Building the measurement before the generator meant we could tell improvement from noise, which is the difference between tuning a system and guessing at one.

The other decision was billing granularity. Charging per element as it generates, and charging nothing for incomplete results, removed the trust problem that kills pay-as-you-go AI products. It was a two-day piece of work with more commercial effect than any feature.

Where it landed

One input produces on-brand output across more than 125 pre-seeded channels in about forty-five seconds, with unlimited custom channels and optional webhooks on completion.