Jessie Mongeon (Dfinity): Leveraging AI for Efficient Technical Documentation and Routine Maintenance

This talk was presented at the AI The Docs 2025 online conference. We are thrilled to share the recording and the summary with you. 

Visit the talk summary page to see all of the presentations from the conference.

presentation

Jessie Mongeon (Senior Technical Writer at DFINITY)

Summary

How can AI support practical documentation tasks such as ensuring consistency, fixing formats, and exposing gaps? 

In her presentation, Jessie Mongeon (Senior Technical Writer at DFINITY) highlights, not just AI for content creation, but AI for maintaining and improving existing docs. She emphasizes specific-purpose chatbots trained on small, targeted corpora rather than broad LLMs. 

What makes documentation effective? 

• Accurate and up to date
• Consistent in tone and format
• Concise but meaningful
• Organized and easy to navigate 

AI applications for documentation 

• Specific-purpose chatbots
  • Reveal gaps (frequent uncertain answers = missing coverage).
  • Surface outdated sources (bot pulling from forums/Discord instead of docs).
  • Test discoverability (if correct info exists but isn’t easily found).
  • Provide onboarding examples (e.g., strong for “deploy smart contract to ICP,” weak for unsupported languages like Ruby, which should be explicitly flagged).
• Mass text replacement and updates
  • More powerful than find-and-replace.
  • Example: changing “alpha” → “beta” across APIs, appending /docv1 to URLs, updating tone across multiple files.
• Style guide compliance
  • Train AI on the style guide alone.
  • Ensures spelling, capitalization, formatting, and tone consistency.
  • Especially valuable for open-source contributions with varied conventions.
• Complex format fixes
  • Goes beyond regex or simple scripts. Examples: Bullet points with * only on tutorial pages. Hyperlink all mentions of “chain fusion” to a marketing page. Add CSS to center images on guide pages but not others.
• AI for routine maintenance
  • CI jobs triggered on every docs update.
  • Scheduled scripts (nightly/weekly).
  • Ad hoc scripts for new releases. 

AI can write these scripts, lowering barriers for non-programmers. 

• Common automations:
  • Check for broken links (404s).
  • Flag non–style compliant docs. List docs untouched for 30+ days. 

Key takeaway: Targeted, doc-trained AI agents can expose gaps, ensure consistency, and automate complex updates. This shifts maintenance from manual, error-prone work to AI-assisted workflows, making documentation more reliable and sustainable.