DocuSync

Companies spend billions answering customer questions; Chatbots are only one part of a comprehensive communication strategy. DocuSync aims to fix the rest, starting with product documentation.

Comment

Inspiration

Companies spend billions per year educating customers about their products, but their product documentation is often out-of-date, inaccurate, or ambiguous.

What it does

DocuSync uses LLMs to review customer feedback and other sources to ensure your product documentation is up-to-date, accurate, and unambiguous.

How we built it

We used OpenAI’s GPT API to find mistakes in the Gmail Help Center, using posts in the Gmail Community forum.

Challenges we ran into

  1. Variability in GPT’s responses makes it difficult to provide reliable output at times. However, this can be circumvented by tweaks to GPT's temperature settings.
  2. The token limit makes it difficult to process text at scale.

Accomplishments that we're proud of

  1. We succeeded at getting GPT to update a draft of Gmail’s Help Center based on a Gmail Community forum post.
  2. We identified many inaccuracies in the Gmail Help Center.

What we learned

  1. Variability in GPT’s responses makes it difficult to provide reliable output.
  2. The token limit makes it difficult to process text at scale.
  3. The OpenAI API intermittently fails to respond (mainly due to server load on their end)

What's next for DocuSync

  1. Additional tuning to improve the system’s ability to detect inaccuracies and suggest improvements
  2. Scale the system to run on larger help centers
  3. Use code changes to update the external and internal product documentation.

Built With

Share this project:

Updates