Teamwork Graph: Feature Suggestion

[Doods Perea]

Dear Atlassian TWG innovators - 

It started with a question:  Does Teamwork Graph analyze the metadata of information it collects from connected apps?

TWG is a unified data layer that maps mined data.  It extracts, maps, and indexes collected data into a common schema.  It does not perform data quality validation or cleansing on source content.

I recommend that Atlassian re-engineers TWG from being just an integration layer/data aggregator into a "data quality governance" layer, wherein organizations are presented with insights and recommendations on mined data from connected apps.  Start small - from its own ecosystem (Jira, JSM, Confluence, Assets, etc.) where Atlassian has the deepest schema knowledge, then expand outward.

TWG should help customers not only connect their information, but also understand its quality, completeness, freshness, and operational reliability.

I thought of sending this recommendation after writing my comment on a LinkedIn TWG-related post: Post | LinkedIn.

Doods

[Viswanathan Ramachandran]

Hi @Doods Perea 

Valuable post and thoughtful comment. You can post this in Teamwork collection group

or in teamwork lab for improved ways of working. Better flag to Atlassian support and support team can raise internally as feature request. 

Given that, here is my feedback and i have posted on Linkedin too
To turn Teamwork Graph from a passive indexing layer into a self-healing context engine, here are three high-impact features I'd love to see on the roadmap:

1. Agentic Infill: Let Rovo auto-generate context (like drafting ADRs) so humans don't have to fill out manual fields.
2. Node Decay & Freshness: Add dynamic confidence scores to graph edges. If a Confluence doc hasn't touched active code in 180 days, decay its weight so Rovo doesn't ground answers in obsolete data.
3. Context Drift Guardrails: Introduce checks in Bitbucket pipelines that flag when code changes diverge from linked architecture nodes.

Solved context hygiene = unstoppable agentic workflows. 

[Mia Tamm _Simpleasyty_]

Hey @Doods Perea this is the part that interests me most. Once metadata becomes an input for AI, completeness and freshness stop being just governance concerns — they directly affect the quality of the answers.

A graph can be perfectly connected and still propagate stale or inconsistent context. I’d love to see things like owner, review date, freshness and required metadata treated as first-class signals rather than something teams have to bolt on later.

[Greg D]

I don't know... idea is interesting and valuable, but security teams have a hard enough time allowing connectors in the first place. Any indication that something is being done with the metadata probably gets the connectors turned off all together in most places (if they were even allowed on in the first place). Security is worried of the unknown and of the vector being bigger if anything were to go wrong.

And I have heard others not want to consolidate for other various reasons.

Do you all have any thoughts on that?