AI Lens GovernanceCopilot agents and Power Automate, one tenant
DEMO · FICTIONAL TENANT
Organization · AI READY

Every AI agent and flow at AI READY, on one page.

A 90-day governance run over Microsoft 365 Copilot agents, Copilot Studio and Power Automate: who built what, who uses it, what breaks and what needs an owner. Collected read-only, built offline.

Open the management dashboard SHAREABLE Full dashboard CONFIDENTIAL
Governance score

At risk

Share of agents and flows without findings. One High finding counts fully, Medium half, Low 0.15.

Run 2026-11-02 · look-back 90 days
Previous run 2026-10-03
28agents · 27 active in 90 days
110cloud flows · 73 ran in the period
4,768flow runs · 13.1% failed
3,083agent interactions by 208 people
327findings · 78 high

Where to look first

Findings of this run by type. Each one names the agent or flow, its owner and the owner's manager, so it can be assigned.

All 327 findings

66 flows failing right now

Recent flow failures is the largest High finding. 13.1% of 4,768 runs failed. High findings cluster in R&D, Customer Success and Data & AI.

12 assets nobody can maintain

10 agents and flows have no owner and 2 belong to a disabled account. They keep running until they break.

7 of 9 environments without DLP

Only the two production environments are covered by a DLP policy, so any connector can be used in the other seven.

8 agents open to the whole tenant

Anyone at AI READY can use them. Check that the knowledge behind them is meant for everyone.

Since the last run

Compared with the run of 3 October 2026. The tool keeps a snapshot of every run and compares automatically.

Changes page
Metric3 Oct2 NovChange
Flow runs4,6534,768+115
Failed flow runs589625+36
Flows with failures6466+2
High findings7678+2
Total findings321327+6
Active agents2827−1
Assets without owner10100
  • New
    New Pilot AgentAgent · Customer Success Prod
  • New
    New approval flowFlow · Default environment
  • Failing
    Inventory threshold alertFlow · Default · started failing since last run
  • Stopped
    Backup SharePoint listFlow · Default · state changed
  • Removed
    HR Policy AssistantAgent · Sales Sandbox
  • Removed
    Lead assignmentFlow · Customer Success Prod

Two dashboards, two audiences

Both are single HTML files that open offline. Same data, different privacy rules.

Management dashboard

For leadership, department heads, the AI steering group
  • Inventory, findings, trends and manager action lists
  • No end-user names: usage is shown as counts per department
  • Any group smaller than 5 people is shown as <5
  • Owners of agents and flows are named, because they are accountable
Open · 2 Nov3 Oct
CONFIDENTIAL

Full dashboard

For the governance team and IT administrators
  • Everything in the management view, down to each user and each run
  • Every flow run with status, duration and error; every agent interaction
  • Filters by department, manager, environment, owner, model and sharing
  • Export of the current view; raw datasets stay as CSV next to it
Open · 2 Nov3 Oct

Runs

Each run gets its own dated folder. The latest one is compared with the one before it.

RunStatusAgentsFlowsFindingsDashboards
2026-11-02Completed · latest28110327
ManagementFull
2026-10-03Completed · baseline28110321
ManagementFull

How a run works

Three stages, run from a PowerShell window on one admin workstation. Saved progress survives a stop: a resumed run does not collect saved stages again.

Sign in

Microsoft Graph, Power Platform admin and optional SharePoint, interactive or app-only.

Inventory

Directory, agent registry, Copilot Studio agents, cloud flows, owners, connectors, DLP.

Run history

90 days of flow runs per environment, with time-outs, retries and checkpoints.

Audit log

Copilot interactions from the unified audit log, matched to each agent.

Build

Findings, manager action lists, trends and both dashboards, all offline.

Read-only: the tool only reads. It sends GET requests plus read-type queries (sign-in token, $batch, audit search) and changes nothing in the tenant.