Quikest

Quikest is a research technology company in Austin, Texas. Their founder described a specific problem: in the US, getting a research protocol written and approved typically takes three to four months, and the delay is not the scientific review, it is writing and formatting the protocol document itself.

While a document sits half finished, a competing team can publish the same idea first. There was no product yet to fix this, just the problem and a clear sense of exactly where the time was going.

We built the founding MVP from an empty repository. At its center is a model fine-tuned to generate research protocols to industry standards, rather than a general assistant asked to try its best.

Researchers could work on the same document at the same time instead of emailing versions back and forth, and a library of templates meant a write-up started in an approvable format instead of a blank page.

It is built on Next.js, the Vercel AI SDK and the OpenAI API.

What took three to four months to get ready was ready in about a week. That is the client’s own account: the three-to-four-month figure is how long US protocol approval normally takes end to end, and the one-week figure is the drafting time the platform delivered. We do not have a usage count from this period, and we are not claiming one.

After the MVP shipped, the client changed the product’s direction three times. The honest reason is worth saying rather than hiding: a large part of what this product did, generating a document from a prompt, can now be done natively by general AI assistants like Claude, ChatGPT and Gemini. The core value was generic generation, and generic generation is exactly the kind of work that gets commoditised fastest.

Looking back, the MVP did what it was built to do. What changed afterward was the market it was built for, not the build itself. It is part of why we no longer sell generic AI generation as the offer: the harder, more durable problem is putting a model safely over a company’s own private data inside their own product, which is a different job and the one we focus on now.

Scope note: we built the research and protocol MVP. A later, persona-based version of the product was built by others.