How to Automate Sprint Rituals With Gemini Enterprise

Standup is fifteen minutes of people re-reading their own Jira tickets out loud. Backlog grooming means someone burning an hour turning forty ticket updates into three bullet points nobody double-checks. Automating sprint rituals with Gemini Enterprise, Google Cloud's agent platform for the workspace tools you already run on, means an agent does the reading and the summarizing, and your team just shows up to react to it. Here's what you'll have by the end: a scheduled agent, built in Gemini Enterprise's no-code Agent Designer, that pulls yesterday's Jira activity and posts a standup draft to Google Chat before your first meeting.

What you need before you start

You need a Gemini Enterprise license, the roles/discoveryengine.editor IAM role on the project (required to create a data store), and admin access to connect Jira as a data source. If your team runs standup in Slack or Microsoft Teams instead of Google Chat, the same steps apply, just swap the connector in step 3.

Step 1: Pick the ritual worth automating first

Don't start with sprint planning, it needs judgment calls an agent shouldn't make alone. Standup notes and backlog summaries are read-and-report tasks: pull structured data, write it up, no decisions required. That makes them the safest first target and the easiest to trust once you see the output for a week.

Step 2: Create the agent in Gemini Enterprise Agent Builder

Open Agent Designer from the Gemini Enterprise app and create a new chat agent. You can describe it in plain language and the builder drafts the workflow on the interactive Flow canvas for you to adjust:

Create a chat agent named "Sprint Pulse" that reads issues updated in the
last 24 hours on our Jira board, plus any messages posted in the
#daily-standup Google Chat space, and drafts a standup summary organized
as: Shipped, In Progress, Blocked.

This becomes your starting point on the canvas. You'll refine the instructions in step 4.

Step 3: Connect Jira and Google Chat as data sources

Gemini Enterprise reads external tools through connectors that sync into a data store. Grant yourself the editor role first if you haven't already:

gcloud projects add-iam-policy-binding <your-project-id> \
  --member="user:<your-email>" \
  --role="roles/discoveryengine.editor"

Then set up a Jira Cloud data store (OAuth or an API token, configured in your Atlassian admin console) and a Google Chat data store, then attach both to your Sprint Pulse agent. Scope the Jira connector to one project or board; indexing your whole Jira instance for a standup bot is slow and mostly wasted.

Step 4: Write the instructions that turn tickets into a standup update

Generic instructions produce generic summaries. Be specific about format and tone:

When triggered, summarize issues updated in the last 24 hours on board
"SPRINT". Group by status: Done, In Progress, Blocked. For each issue,
include the ticket key, title, and assignee. Flag anything blocked for
more than 2 days at the top. Keep the whole summary under 200 words and
write it in plain sentences, not raw ticket dumps.

Run it a few times from the test panel before you trust it near a channel your whole team reads.

Step 5: Add a daily schedule trigger

In the agent builder canvas, open the Schedule tab and click + Add schedule. Pick a recurrence (weekdays) and a time that lands before standup, say 7:45 a.m. This works for a straight chat agent; execution schedules aren't supported on workflow-type agents, so keep this one as a chat agent, not a multi-step workflow, if you want it to run itself.

Step 6: Post the summary where the team already is

Point the agent's output at your #daily-standup Google Chat space instead of leaving it sitting in a chat history nobody opens. The same pattern works for a weekly backlog summary posted to your planning channel, just change the trigger to weekly and the instructions to cover the whole backlog instead of the last 24 hours.

Step 7: Keep a human in the loop before it ships

For anything that posts publicly, add an approval step on the Flow canvas so a lead reviews the draft before it goes out, at least for the first few weeks. Workflows in Agent Designer support mixing AI steps with human review steps, so this doesn't require a separate tool.

Gotchas

• Schedules only work on non-workflow chat agents. If you built a multi-step workflow, you'll need a different trigger, like an event.
• Jira connector auth needs an Atlassian admin, not just a Google Cloud admin. Loop them in early.
• Scope your data store narrowly. A board-wide sync is slower to set up and pulls in noise your summary doesn't need.
• Test output for a week in a private space before it posts anywhere your whole team sees by default.

The fifteen minutes standup takes isn't the real cost, it's the context-switch of stopping to write it. Handing that to an agent that already has the data is a small win, but it's the kind that compounds every single sprint.