NewIntroducing Agent Crew for Microsoft Teams

Share intelligence, infinitely.

ShareAI builds AI teammates that work inside the tools your team already uses. They understand the full story of a project, connect knowledge scattered across messages, work items and documents, and raise problems before they turn into incidents. People always make the final call.

Try the demo Meet Agent Crew

# DeliveryNorthstar Commerce Demo

PM is watching Sprint 24 · 3 roles consulted

PMAGENT CREW09:02

I found a delivery risk while checking this sprint. It should be resolved today.

High#142 Refund flow has no acceptance criteria

Two days left in the sprint, and QE can't start writing test cases
Partial-refund rules from the design review aren't in the work item yet

ADO #142Design Review threadQE input

Update #142 and notify owner? EditApprove #142 updated · owner notified

Waiting for Olivia Grant to reviewApproved by Olivia Grant · 09:04

Ethan Brooks09:05

@Junior Dev Once the criteria are in, which parts of the refund module will we need to change?

5 rolesPM, QE, Developer, UX and Data Analyst

In TeamsWorks in the channels your team already uses

Human-approvedNothing is sent or changed without review

SourcedEvery answer links back to where it came from

Our mission

Every team deserves an AI team that knows its project.

Most AI assistants work one question at a time. You explain the background again every time, and the conversation ends when the answer does. We are building teammates that stay with the team, remember the project, and speak up on their own.

Shared context

AI and people see the same project: discussions, requirements, work items and decisions. Every answer can be traced back to its source.

Humans decide

AI can draft, suggest and remind. Before it sends a message or changes a work item, a person reviews and approves it.

Work where you work

No new platform to adopt. AI teammates show up in the channels your team already uses, and an @mention is all it takes.

Flagship product

Agent Crew

An AI project team that works inside Microsoft Teams. It understands project context, spots delivery risks early, coordinates five specialist roles, and drives follow-up actions under human review.

PM

Project Manager

Tracks progress and scope, checks the sprint for risks, and rounds up the decisions people need to make.

QE

Quality Engineer

Reviews acceptance criteria and test coverage, and flags requirements that can't be tested yet.

Junior Dev

Developer

Explains code and the impact of changes, and helps new teammates get up to speed on a module.

UX

Designer

Checks work against design review decisions and points out where the build and the design differ.

Data Analyst

Analyst

Answers questions with data and explains what a change in a metric may mean.

Governed by design

Agents propose. Your team approves.

One review queueDrafted messages and work item changes wait for the right approver.
Drafts that know when they're staleIf a source changes, the draft is flagged before anyone approves it.
Separate projects stay separateMembers of one project can't see another project's content.

My reviewsOlivia Grant Demo

D-004 · PMUpdate #142 acceptance criteria

Needs review

D-003 · QENotify owner: test plan blocked

Needs review

D-002 · PMSprint risk summary

Source changed

D-001 · UXFlag checkout copy mismatch

Approved

How it works

From risk to action, with a human in the loop.

Agents do the reading, watching and drafting. The step that changes your systems belongs to a person.

AGENT
Understand
Reads channel discussions, work items and design reviews to build a lasting memory of the project.
AGENT
Watch
Checks progress, gaps and dependencies on a schedule, without waiting to be asked.
AGENTS
Consult
The PM leads, and QE, Developer and other roles add their own view of the problem.
HUMAN
Review
The owner approves, edits or rejects the drafted alert or work item change.
AGENT
Act
The message is sent and the work item updated, with sources and approval on record.

Why Agent Crew

Not another chat box. A project team.

General assistants are good at answering questions. Agent Crew is built to carry part of the delivery work.

	General AI assistant	Agent Crew
Memory	Starts from zero every conversation	Remembers the project's requirements, decisions and history
Initiative	Waits to be asked	Checks the project on its own and raises risks in the channel
Perspective	One generic answer	PM, QE, Developer, UX and Analyst views on the same question
Evidence	Sources are often unclear	Links each answer to the post or work item it came from
Changes	You copy the answer and act on it yourself	Drafts the change, waits for approval, then carries it out

Capabilities and roadmap

What you can try today, and what comes next.

Agent Crew is an interactive demo prototype today. The left column is what you can try in it now. The right column is what we plan to build next.

In the demo prototype

Available now

@mention a role in a Teams-style channelPM, QE, Junior Dev, UX and Analyst each answer
Proactive risk alertsThe PM checks the sprint and raises risks in the channel
Work item proposals with approvalNothing is written until the owner approves
Draft expiry and stale-source warningsWhen a source changes, older drafts are marked out of date
Answers with sources and reasoningJump to the original post, or expand each role's input
Data separation between projectsMembers of another project can't see this project's content

Planned

Coming next

Connect to Microsoft Teams and Azure DevOpsLive channels and work items instead of the demo environment
Auto-send trust rulesLow-risk alerts go out automatically once approved several times
Team setup wizardConfigure check rules, active roles and approvers
Risk follow-up and escalationFollow up again when nobody acts on an alert
Answer feedbackRate each answer or flag a wrong one, to keep improving quality

The mark

One stroke for Share, one loop that never ends.

S for Share

The first letter of ShareAI, and the act of sharing itself.

∞

Infinity for endless sharing

The line loops back on itself, the way knowledge should move through a team: what one person knows, everyone can use.

···

Three nodes for people, AI and knowledge

People ask and decide, AI understands and acts, and project knowledge connects the two.

See it in action

Watch Agent Crew catch a risk before it ships.

A welcome card guides you through the main story the first time you open the demo. It takes about four minutes.

Open the demo