Using AI to produce summaries is not an invention of generative AI. What is new is the ability to combine this pattern with others to making performing a mundane task easier and faster (or make it go away altogether).

An average person can think about how much time they spend doing routine tasks: Catching up on meetings you missed, prepping for conversations, reading documents, etc. The summary pattern provides a shortcut for users to get the essence across all of the inputs into their day, and focus on what really matters.

  • Combined with citations, summaries let you capture key takeaways from a long document and then go deeper into the content that are relevant to you
  • Combined with synthesis, summaries can take multiple inputs such as an email thread and capture the primary takeaways
  • Combined with workflows, summaries can be added automatically to specific documents or other locations for easy retrieval

Be mindful to label summaries clearly using a disclosure badge, similar to how Productboard notes the AI Summary of any long insights. Also, be aware of the possibility for information to be missed, or inaccurately reported by the AI. Citations and sources are your best friend with this pattern because they empower the user and help them engage deeper in a more meaningful way.

Details and variations

  • Summaries can be auto-created or created on demand
  • If summarizing multiple references synthesized together, show citations and other paths back to the sources themselves
  • Let users adjust what sources to be included in the summary
  • Label the summary clearly with a badge, and ensure users know that some data could be missed



No action needed

While it's helpful to offer nudges to let users know the AI can help them summarize content down, this is a low risk pattern that can be introduced up front. Video transcript apps, insight apps, news app - all have started to introduce the concept of AI Summary by default. Not all prompts require the users to write something or push a button.

Potential risks

Don't over simplify

AI summaries can make it easy to get to the information that matters, but they don't replace the information themselves. Find paths back to the original information through citations and sources to help users go deeper where it matters, and not lose the context of what the AI is reporting.

Use when:
The user will need to consume a bunch of information to get to the essence of a topic, and the AI can help guide them to uncover what matters


Slack is rolling out the ability to summarize a thread from within the interface
Grain and other video recorders include an AI-generated summary next to the transcript, and often timestamped
Productboard also summarizes multiple insights under a client record or story where they are aggregated
Salesforce Einstein shows another example of inline summarization, available in context while the user is looking at the customer record
Grammarly can summarize into a shorter version or into an outline
Notion allows you to embed cards to summarize the data on the page with a single command
Jira makes summaries personal to the user, and can include links to other sources to set up next actions
No items found.