What an AI agent can do
Once connected, an AI assistant can help with your work items, change sets, and configuration. It acts as you, and never with more access than you have.
A connected assistant works through the same actions you can take yourself. Whether it can change anything at all is set separately from whether it can look, and turning off access stops it immediately. Here's what it can help with once you've given it access.
Reading comes first
With AI access on and nothing more, the assistant can only look. It can read your configuration and its history, list change sets, work items, pipelines, and runs, and check what's in an environment. Everything below that creates or edits something needs an admin to allow changes for the workspace, which is off to begin with.
Work items
The assistant can list your work items, create new ones, and update them, including their acceptance criteria and which board column they sit in. You can describe what you want in plain language and have the assistant capture it as a work item with a checkable definition of done.
Change sets and configuration
The assistant can help you build and manage change sets:
- Look at an environment. Capture what's in an environment and see the items it holds.
- Gather a change set. Group the items you want into a change set, with a description. Mention a work item's number like #12 and it links automatically.
- Submit it. Add the change set to your configuration, going through approval if your workspace requires it.
- Read your configuration and its history. See what's tracked, how items relate, and which change sets touched a given item over time.
- Follow delivery. Look at your pipelines, releases, and the runs that carry a change from one environment to the next.
Working in a connected Aprimo instance
For any Aprimo instance you've connected, the assistant can help build and adjust configuration there, using your Aprimo account and permissions. You describe a requirement, and the assistant can both make the change in the instance and capture it as a change set linked to the work item.
Every one of those actions is sorted in advance into two piles: the ones that only look something up, and the ones that change it. An instance an admin connected read-only runs the first pile and refuses the second, so pointing an assistant at production to investigate something stays a safe thing to do. A handful of the changing ones can't be taken back, and those are flagged so your assistant can ask you before it runs one.
A reviewer can tell what it proposed
A change set the assistant put together carries the name of the person it's acting for, because that's who authorized it. It's also marked as put together by AI on the review screen. That's the difference between reading the changes properly and just approving, so it's worth knowing which one you're looking at.
What it can't do
- Approve a change set, whether it put that one together or someone else did.
- Turn down someone else's change set.
- Send a release out, start a run again, stop one that's under way, or put an environment back to an earlier version.
- Create or replace an access key, which would keep working long after the conversation ends.
- Change anything in a workspace where an admin has allowed AI to read but not to change.
- Make changes in an Aprimo instance that's connected read-only.
- Reach workspaces you're not a member of, or ones with AI access turned off.
- Skip the review and delivery your workspace has set up.
Mochi manages your Aprimo configuration, not your content, assets, or workflows, so the assistant works with configuration too. See what Mochi tracks.