Following the recent launch of Insight 2.0, we have some exciting AI Chat upgrades and tool updates this month.
Some of these updates are based on your feedback, and others are new capabilities made possible by more advanced AI models.
AI Chat Memory
You can ask AI Chat to remember information and preferences, making chat responses more personalized for you. Chat memory is private, not shared with your team.
How it works: Your saved memories are loaded at the start of every conversation. The AI uses your memories as context whenever they are relevant to the conversation.
Data storage: Your chat memories are stored in Insight, not with any AI providers, and never used to train AI.
Save a memory: Simply tell the AI what you want it to remember, and AI Chat will confirm it’s been saved. For example:
- Remember that I’m a children’s book editor at XYZ Press.
- Remember that my company publishes university textbooks.
- Remember that our main markets are Germany and Austria.
- Remember never to use the phrase “dive into”.
- Remember that I prefer bullet points over long paragraphs.
- Remember that our fiscal year runs from April to March.
View your memories: Ask the AI to show your saved memories at any time. Use phrases like: “Show my memories” or “What do you remember about me?“
Update or remove memories: Describe the change you want to make. For example:
- Forget that I’m a children’s book editor.
- Update my memory about our company — we now also publish general nonfiction.
- Add to my memory about our markets — we also sell in Australia and New Zealand.
- Delete my memory about phrases you shouldn’t use.
- Delete all my memories.
Tip: It’s helpful to first ask AI Chat to show your saved memories, then ask for changes.
AI-Powered Chat History Search
You can now search your chat history from inside AI Chat.
Similar to internal document search, this AI-powered search looks for content by meaning. It responds with links to specific conversations and a brief overview.
Try it: Go to AI Chat and enter prompts like “Search my chat history for conversations about proofreading” or “Find my chat conversations about the book Mrs. Dalloway”.
New LLMs
Both Anthropic and Google have released new LLMs that are better at complex, multi-step tasks. These are now available for you in AI Chat (and in background processes across Insight).
- Sonnet 4.6: Better at instruction following, planning (multiple steps and multiple tools), and general reasoning
- Gemini 3.1: Improved reasoning and token efficiency, and better performance on multi-step tasks
Improvements to Visual Analysis
With the latest AI models, we’ve been able to improve the accuracy of vision mode across our document tools and document processing (for example, understanding a graphic novel).
In particular, our Final PDF Check in the Scan tool gives you more consistent results and fewer false positives.
Try it:
- From a PDF document, click Actions > Document tools > Scan Document
- Choose “Final PDF Check” from the template list
- Click “Start Scan”
Proofreading Tool Updates
Rulesets, or collections of individual proofreading rules, are now the primary way to work with this tool. Rather than selecting individual rules, save your selections as a ruleset and then apply it to your document.
Read the guide: Getting Started with Proofreading
We’ve also added some default rulesets, which you can use as a starting point for creating your own.
A few more proofreading tips:
- Duplicate and customize one of the default rulesets. Open the rule editor and click the copy button on a ruleset. Then remove rules you don’t need and add your own.
- Look at the default rules for guidance. Click “View full rule” on any system rule to see how it’s written. Effective rules state one clear instruction followed by a few supporting bullet points.
Create separate rulesets for different needs. Build different rulesets for your imprints, book types, and special projects.