ProspectAI has been built to streamline online searches and report creation for prospect research. Without getting into too much detail, here's how that works:
- When you enter a prospect's details (name, location, etc), the tool first uses Google APIs to make a series of online searches. These are the kinds of things you might search for manually (e.g., "John Smith" AND "gifts").
- The tool the aggregates the top few pages of all of those results, and uses a different software packages and APIs to figure out what the content is about (e.g technology for reading PDFs). This step is analogous to a prospect researcher manually reading all of those sources.
- Next, the tool uses large language models (right now it uses OpenAI's API) to help with identity resolution - this basically means "of all these websites/PDFs that refer to John Smith, how many unique people named "John Smith" are there"? This is one of the things that modern large language models are reasonably good at (but not perfect, especially if the sources are vague).
- Once the tool has identified the primary "John Smith" - the one that you were ostensibly looking for - it again uses LLMs (right now this is OpenAI's API) to summarize the sources and construct a structured report based on the source material.
- There are a number of other discrete steps happening in parallel with the above to provide great accuracy and context to the reports, like looking through many other public data sources, but the above is the basic process.