Persistent Memory for AI Models
planned
Dustin Kuntz
It would greatly improve productivity if AI models in ClickUp could remember my conversation preferences—such as preferred response format, tone, and level of detail—across sessions. This would save time by eliminating the need to repeat instructions and ensure a more personalized, consistent AI experience for ongoing projects.
Log In
Mike Certoma
I’ll second tone and style for sure.
Things like company-specific rules, internal terminology, approval flows, definitions of “done,” how workflows are structured, how we communicate with clients vs internally, and even what not to do.
Right now, that context lives in SOPs, tribal knowledge, or people’s heads. The AI has no awareness of it unless it’s re-prompted every time, which quickly becomes friction, especially for ongoing work.
Even a lightweight, user-defined “company memory” the AI could reference would significantly improve accuracy, consistency, and trust in the output. Something conceptually similar to selectable memory banks, like NotebookLM, would be a great model.
A related issue is context depth. Even when this information already exists inside ClickUp, the AI often struggles to reliably reference or reason across it. A strong baseline solution would be allowing LLMs to source from defined lists or documents as a dedicated memory bank, with enough depth and consistency to make that context dependable rather than hit-or-miss.
E
Erra Faye Zabat
Hey, everyone! We've updated the status of this request to reflect its roadmap status. The Product Manager will provide updates on this post when they have new details to share! For more details about our statuses, check out this post!
E
Erra Faye Zabat
marked this post as
planned
Alex Omeyer (ClickUp)
Dustin Kuntz totally agree — I'm itching to build this. It hasn't quite hit the top of the prio list but we'll get to it!
In the meantime, any features / use cases that are particularly important to you? Sounds like tone and general response styles are the main ones.
Dustin Kuntz
Alex Omeyer (ClickUp) Yes, tone and general response styles are the main ones for sure. However, I've seen other's request it, and I agree that having some kind of memory that is "company wide" would also be helpful. Something that says "company wide, this is how we do invoices" that the AI would be able to reference without being told with prompts, to refine it's answers to be more accurate in the context of our own company. Essentially, somewhere where an admin could put SOPs for any number of things that the AI could draw context for is what I'm envisioning.
Alex Omeyer (ClickUp)
Dustin Kuntz: Ok understood, this is a very cool use case. I've made a note to consider this when we work on the memory features — can't wait!
Alex Omeyer (ClickUp)
marked this post as
considering