I have a workspace with 100 users, 15 agents, and recently increased from 10 memory slots to 46 memory slots per user. During the creation of the new memory slots, I messed up and DELETED ALL OF MY SUBSCRIBERS MEMORIES!!!
For a few days I was mad, embarrassed, frustrated, sad, etc… if you’ve done it, you know…
Then I had an idea!
I asked Wingman if it could scan the chat history for a specific user across all agents and use that to fill in the memory slots. It could.
Here’s what it actually did:
It pulled session history from every agent for a target user. It mapped what was already stored in memory so it wouldn’t overwrite anything. It analyzed the conversations to extract rank, voice and tone, DMO preferences, family context, mindset blocks, faith study focus, dietary patterns, outreach style, business goals — anything that had been mentioned naturally in conversation. Then it wrote only to the empty slots, leaving the already-filled ones alone.
For my own profile it filled 18 missing slots in one pass. Then I asked it to do the same for all active users and it filled 99 total gaps across the workspace in about 10 minutes.
The most important thing I noticed: it matched the data to the right person. I asked it to verify a specific user’s memories against mine to make sure nothing crossed over. Zero identical values. Every slot was sourced from that person’s actual conversations.
If you’re running a multi-agent workspace where memories are your personalization engine, this is worth exploring. The data your users have already given you in conversation is often more complete than you realize. Wingman can surface it.
Worth noting: Wingman also did a pre-scan to identify which users had enough conversation history to make the rebuild worthwhile, so it wasn’t writing generic or fabricated data for low-activity accounts.
Happy to answer questions if anyone wants to try this in their own workspace.