AI agents and AI transcription

The problem

Outfield users spend their day on CRM records. They create contacts and deals, search for records and clean up duplicates. They also turn spoken notes into written notes and forms. [TODO: confirm that this is how you describe the need]

What I built

I prototyped and built AI agents with RubyLLM. The agents create contacts and deals, search records and merge duplicates. They do all of this through the platform's REST API.

I also built an AI transcriber. It turns audio into written notes and fills in forms.

Decisions

Act only through the REST API. The agents use the same REST API as the rest of the product. [TODO: why? For example, the same rules and checks as a normal user.]

Prototype first. I built quick prototypes of AI features, such as summaries of team activity. I checked each one against the real limits of the product, and only the strongest ideas became working features.

Result

Users can run core CRM tasks with plain language. [TODO: a result you can share, for example usage or time saved]