Compose.ai, which built its own language model to help folks write faster, landed a $2.1 million seed round led by Craft Ventures. The company is building the capability for its AI-powered backend to learn your voice, imbibe context to help provide better responses, and, in time, absorb a company’s larger voice to help align its aggregate writing output.
- An innovative future: Co-founders Landon Sanford and Michael Shuffett said that Compose.ai believes that average folks won’t type every word they write in five years. They want to bring that future to more people through the Compose.ai Chrome extension, which hopefully workers can access without having to get corporate permission.
- A multi-tier project: Sanford and Shuffett described their language algorithm as a multi-tier product. Its first tier is learning from reading the internet itself, learning English. The second tier deals with specific domains, like email. The third tier learns a user’s voice, and, in time, a fourth tier will deal with a company’s generalized approach to language.
- Tuned up finance: What Compose is building isn’t technically simple, and the company has interesting work ahead of it to get its economics properly tuned. The company’s founders said that the personalization work that its product will execute for customers can be expensive if done incorrectly; the pair said that they could make a future, $10/month plan attractive economically, but if done in a “naive” fashion, Compose.ai could spend $500 in compute costs to support the same account.
- An Aid for the future: The ability for a company to provide its workers with a shared language model that could offer linguistic and word-choice preferences as they write is an interesting concept. The idea of such a strong, centrally held tone lands somewhere between helpful and intellectually rigid. But lots of folks don’t want to put their own spin on writing; many people don’t like writing at all. So perhaps having more central support won’t be onerous to most — but rather a time-saving hack.
- The road ahead: Sanford and Shuffett said that the company intends to stay small, hiring a few experienced engineering and machine-learning staff to help flesh out its product team. Its founders said that they are starting to talk to the corporate world a bit more than before and are planning on launching a paid service toward the end of the quarter. That could mean early revenue, extending the startup’s runway substantially.