How to build an AI project manager with Sai

An AI project manager keeps projects moving between meetings: it writes the standup digest, turns meeting notes into tickets, triages new bugs, flags blocked and at-risk work, and sends the weekly status report. This guide builds one with Sai, step by step, using Linear or Jira, Notion, Slack, GitHub and Google Calendar, with every change to scope, priority or owners waiting for a person.

Tools: Linear, Jira, Notion, Slack, Google Calendar, Google Sheets

TL;DR: the AI project manager you end up with

By the end of this guide you have an AI project manager that does the reporting and follow-up work of running projects, every day, so the team spends meetings on decisions instead of status. Priorities, scope and who works on what stay with people. It runs six routines:

Standup digest. Every weekday before standup, posts what got done, what is blocked and what is at risk, per team.
Meeting notes to tickets. After planning meetings, turns decisions and action items into Linear or Jira tickets with owners.
Bug triage. Labels new bug reports, sets a suggested priority and owner, and links duplicates.
Risk watch. Flags tickets with no movement, scope that grew mid-sprint and deadlines that no longer fit.
Weekly status report. Every Friday, writes the project update for stakeholders in Notion and Slack.
Release notes. Turns merged tickets into customer-facing notes for each release.

A Sai Ready task, Standup digest for the Platform team: two tickets done, one blocked on Stripe test keys, one at risk with two days left and no pull request, and a release review at 2 PM with notes going to Notion — Ready: the standup digest, posted before the team meets.

What is an AI project manager?

An AI project manager is an agent that runs the information side of project management: collecting status from tickets, pull requests and chat, writing digests and reports, turning notes into tickets and chasing what is stuck. It does not set priorities or make trade-offs. It makes sure the people who do have an accurate picture every day.

Sai fits because project status is scattered. The truth sits across Linear or Jira, GitHub, Slack threads, meeting notes in Notion and calendars. Sai reads them in one task, through integrations where they exist and in a browser where they do not, and writes back where the team already looks. For the Notion side in depth, see Sai × Notion.

What your AI project manager will do

Start with the standup digest. It is useful from day one and teaches Sai how your team works.

Workflow	What Sai does	Cadence	Needs your approval
Standup digest	Done, blocked, at risk, today, per team	Weekdays 9:15	No
Meeting notes to tickets	Decisions and action items into tickets with owners	After planning meetings	Yes, before creating
Bug triage	Labels, suggested priority and owner, duplicates	On every new issue	No for labels, yes for priority
Risk watch	Stalled tickets, scope creep, deadline slips	Daily	No, it flags
Weekly status report	Project update in Notion and Slack	Fridays 15:00	Yes, before sharing outside the team
Release notes	Customer-facing notes from merged tickets	Each release	Yes, before publishing

What you need before you start

A Sai account. The Free plan includes one cloud computer for building and testing. Daily routines for several teams work best on Pro. See Plans, usage, and limits.
Connections. Linear or Jira, Notion, Slack, GitHub and Google Calendar in Abilities → Integrations. Anything without an integration can be signed in on Sai’s cloud computer through its approval link.
Your project rules. What each status means, what counts as blocked or at risk, and who owns what.
A home for reports, such as a Status page in Notion and a #project-updates channel.

Step 1: Write your Project Rules skill

This skill is how Sai knows the difference between a ticket that is fine and one that is in trouble.

When to use it
Any task about our projects, sprints, tickets, standups or status reports.

Statuses
- Blocked: waiting on someone outside the assignee for more than 1 business day.
- At risk: estimate left > days left in the sprint, or no commit or PR in 3 days on an In Progress ticket.

Teams
- Platform (Linear team PLAT), Growth (GRO), Mobile (Jira project MOB).

Reports
- Lead with the answer: on track, at risk or off track, and why, in one line.
- Name owners. Link every ticket.

Never
- Change priority, scope, sprint or assignee without my OK.
- Close or delete tickets.
- Message anyone outside the team without my OK.

Sai Abilities → Skills with project management skills: Project Rules, Status Report, Standup Digest, Meeting Notes to Tasks, Release Notes, Bug Triage, Sprint Review and Capacity Check — Abilities → Skills: one skill per job, so each routine follows the same rules.

Step 2: Post the standup digest every morning

Standups get shorter when everyone has already read the status.

Every weekday at 9:15, for each team in my Project Rules skill:
list tickets moved to Done since yesterday, tickets that are Blocked or At risk with the reason,
and today's meetings that affect the sprint from Google Calendar.
Post one message per team in its Slack channel and add it to today's Standup page in Notion.

Before standup, list what each team finished since yesterday in Linear, what is blocked or at risk with the reason, and today’s meetings that affect the sprint. Post one message per team and add it to today’s Standup page in Notion.

3 teams · 46 tickets
12 open PRs
Release review at 2 PM
3 digests posted
Standup · Tue Oct 21

Sai read 46 tickets and 12 pull requests across three teams, posted a digest to each team’s channel with what got done, one blocked ticket and two at risk, and added today’s Standup page in Notion.

Step 3: Turn meeting notes into tickets

Decisions made in a meeting are lost if they never become tickets.

After each meeting called Planning or Review on my calendar, read its notes in Notion.
List each decision and action item with an owner and a due date.
Draft a Linear ticket for each action item in the right team, linked to the notes.
Show me the list before creating anything.

Step 4: Triage new bugs

A bug with a label, a priority suggestion and a duplicate check is half fixed.

On every new issue labeled bug in Linear:
- Add area labels and a suggested priority with a one-line reason.
- Search for duplicates and link them.
- If it came from support, link the tickets and count affected customers.
Ask the team lead before changing priority.

Support teams can feed this from the queue. See the AI customer support agent.

Step 5: Watch for risk every day

Most slips are visible days before they happen. Sai looks every day.

Every day at 16:00, check every In Progress ticket in the current sprint.
Flag tickets that match Blocked or At risk in my Project Rules skill, tickets added to the sprint after it started,
and epics whose remaining estimate no longer fits their target date.
Send me one Slack message with the list and a suggested action for each.

Step 6: Send the weekly status report

The update stakeholders read, written from the tickets, not from memory.

Every Friday at 15:00, write this week's status report on the Status page in Notion:
one line per project (on track, at risk, off track and why), what shipped, what slipped,
risks and decisions needed, and next week's goals. Link every ticket.
Post a five-line summary in #project-updates. Ask me before sharing outside the team.

Write this week’s status report on the Status page in Notion from Linear and Jira: one line per project with on track, at risk or off track and why, what shipped, what slipped, risks and decisions needed. Post a summary in #project-updates.

6 projects · 118 tickets
14 shipped · 5 slipped
Week 43 report
Posted

Sai wrote the week 43 status report in Notion: 4 projects on track, 1 at risk and 1 off track, each with the reason, 14 tickets shipped and 5 slipped, and two decisions needed. The team summary is posted; sharing with stakeholders waits for you.

Sai Routine board for project work: Stakeholder update waiting for approval; on a schedule, the weekly project status report on Fridays at 15:00, triage of new bug reports on every new issue, a standup digest on weekdays at 9:15 and meeting notes to tasks after every planning meeting; paused, a daily Jira cleanup and a capacity check from Linear — Routine: every project routine in one place, with the stakeholder update waiting for your OK.

Step 7: Put your AI project manager on a schedule

Once each task runs well by hand, make it a routine:

Routine	When it runs	What you do
Standup digest	Weekdays 9:15	Read it before standup
Meeting notes to tickets	After planning meetings	Approve the tickets
Bug triage	On every new issue	Confirm priorities
Risk watch	Daily 16:00	Act on flags
Weekly status report	Fridays 15:00	Approve sharing

See Routines and Skills.

Keep your AI project manager safe

Priorities are people’s calls. Sai suggests; leads decide scope, priority and assignments. Put that in the skill.
No closing or deleting. Sai can comment and label, but never closes or deletes tickets.
Scoped access. Connect only the teams and projects Sai needs, and keep private channels out of reach unless you add them.
Status, not surveillance. Report on work, not on individuals’ hours. Agree with the team on what the digest includes.
Ask before sharing outside. Stakeholder and customer-facing updates wait for approval.

Why build your AI project manager on Sai

It reads everything the team writes. Tickets, pull requests, Slack threads, notes and calendars in one task.
Your rules are a paragraph. What counts as blocked or at risk lives in a skill, not in tool-specific automation rules.
It works across tools. Linear for one team and Jira for another is fine; Sai reads both.
It finishes the task. Sai ranks #1 on OSWorld for completing real work on a real computer.

Frequently asked questions

What is an AI project manager?

An AI project manager is an agent that collects project status from tickets, pull requests and chat, writes standup digests and status reports, turns meeting notes into tickets and flags risk. People set priorities and make trade-offs.

Does Sai work with Linear and Jira?

Yes. Sai connects to Linear and Jira through Abilities → Integrations, and can use either in a browser on its own computer for anything the integration does not cover.

Can Sai turn meeting notes into tickets?

Yes. Sai reads the notes, lists decisions and action items with owners and due dates, drafts a ticket for each, and creates them after you approve.

Will Sai change priorities or assign work?

Not without approval. Sai suggests priorities and owners with a reason; a lead approves the change.

Can Sai write the weekly status report?

Yes. Sai writes it from the tickets: one line per project, what shipped and slipped, risks and decisions needed, with links, and posts it where you choose.

How does Sai decide a ticket is at risk?

With the rules in your Project Rules skill, such as remaining estimate versus days left, or no commit or pull request in three days.

Does Sai replace a project manager?

No. Sai does the status collection and follow-up so project managers can spend their time on planning, trade-offs and people.

How much does it cost?

You can build and test every step on the Free plan. Daily routines for several teams usually need Pro at $50 a month. See Plans, usage, and limits.

Related guides

Keep building on the same pieces.

Routines for this role: timesheet reminders and automated client reporting.