← Everyday AI series
Session 07 March 2026 · AI

Custom GPTs & Sharing

With agents and custom GPTs both in our toolkit, the real skill became judgment — knowing which tool to reach for when. This session paired that decision-making with a hands-on return to custom GPTs, and a share-and-tell that turned individual experiments into team-wide know-how.

Which AI, when?

The framing question for the whole session: not "can AI do this?" but "which AI should do this?" We talked through where each option earns its place:

Reach for a Copilot agent when…

  • You need to connect to a knowledge base or source data (our chat license is restricted from doing that)
  • The task is ongoing and data-driven
  • It should live where others can use it

Reach for a custom GPT when…

  • You're shaping how something is produced — tone, format, structure
  • The job is a repeatable transformation
  • You want to package a workflow and share it

Building more complex GPTs

I circled back to ChatGPT and walked through building custom GPTs again — but this time more ambitious than our first pass. Real examples we built:

  • Course → microlearning — a GPT that takes an existing course and breaks it into bite-sized microlearning chunks.
  • On-brand slide decks — a GPT that applies our branding and tone of voice to produce better, more consistent slides.

Share & tell

Then came the part that makes a team smarter, not just an individual: I shared out the GPTs I'd already built so everyone could see the prompts behind them and try them directly. Others shared theirs too, and we opened the floor — who's building what, what's working, what isn't. The session became a genuine exchange, not a lecture.

Key takeaways

  • Match the tool to the job — agents for connected data, custom GPTs for repeatable transformations.
  • Custom GPTs can handle complex, multi-step work like restructuring a course or enforcing brand.
  • Sharing the prompts behind a GPT lets others learn and adapt, not just consume.
  • Share-and-tell compounds — everyone's experiments become everyone's toolkit.

Delivered live to my team and recorded internally for anyone who missed it.

NEXT IN THE SERIES
Session 08 · Scheduling GPTs →