Multi-Agent Automation
for the Audit Floor
A vision for compressing fund-audit workflows with a team of orchestrated AI agents — collecting evidence, populating working papers, and surfacing status — all driven conversationally from Microsoft Teams.
Fund audits still run on manual glue
The work is rigorous, but the coordination around it is stitched together by people moving documents and status between tools.
Evidence lives everywhere
Source documents arrive by email, portals, and shared drives — collected and classified by hand, one file at a time.
Procedures assigned ad-hoc
Who does which procedure is negotiated in meetings and tracked in a spreadsheet that drifts out of date within days.
Exceptions get buried
Anomalies surface late, scattered across inboxes and threads, with no single place that shows what still needs a decision.
One orchestrator, a team of agents
A central orchestrator delegates each part of the audit to a specialist agent, then pulls the results back together — the auditor supervises the flow instead of driving every step by hand.
Orchestrator
Coordinates the agents & talks to auditors in TeamsPBC Intake
Reads the shared mailbox & classifies client files into SharePointWorking Papers
Populates the Excel working papers via Office ScriptsStatus & Reporting
Shows clients where each audit stands in Power BI / Power AppsException Review
Flags anomalies & routes them back to the auditor in TeamsIt all happens inside Microsoft Teams
Conversational control
Auditors talk to the orchestrator in plain language — reclassify a document or ask for status without leaving the chat.
Notifications where they already look
New PBC files, exceptions, and status updates arrive as Teams messages — not another inbox to check.
Native, not a replacement
It plugs into the Microsoft suite auditors already use every day — no new tool to learn, no major disruption.
3 new PBC documents landed in the shared mailbox. Classified in SharePoint: 2 bank confirmations, 1 trial balance.
One file didn't match a known category — I filed it under “To review”.
Move the last one to Legal confirmations.
Reclassified in SharePoint and updated the working paper. The Power BI status is refreshed.
Low-code, pro-code where it counts
The whole vision sits on the Microsoft low-code platform — fast to assemble and easy to hand over — with pro-code (Office Scripts) exactly where the audit logic needs it.
Copilot Studio
Authors and hosts each conversational agent
Microsoft Teams
The interface auditors talk to the agents through
SharePoint
Stores and classifies the Provided-By-Client evidence
Power Automate
Orchestration flows that move work between agents
Power BI / Power Apps
Client-facing dashboards of where each audit stands
Office Scripts
The pro-code that populates the Excel working papers
Audit teams supervising a fleet of agents — not doing the rote work themselves.
That's the direction this concept points at: orchestrated AI handling the heavy lifting, humans owning judgment and review — all inside the Microsoft tools auditors already use, so it makes their day easier instead of adding another system to learn. It's a vision I'm actively exploring in applied AI and low-code systems.