Related posts
See all
Key insights
- Coordination work like status compilation and evidence chasing moves to AI agents, freeing managers to focus on judgment work like exception review and escalation.
- As execution hours shrink, review becomes the visible part of the manager's job and supervision becomes a tracked quality metric alongside realization and utilization.
- Managers who learn to direct AI agents and design their own review checkpoints now will define the version of the role that gets promoted.
Most audit managers know the feeling of losing a Tuesday to status work: chasing aging requests, reconciling drafted workpapers, unblocking staff, following up with clients. None of it is the judgment the firm is paying manager rates for, but all of it has to happen for the engagement to move.
That coordination layer is the first thing agentic AI takes off the plate. AI agents handle evidence intake and status tracking inside the workflow, draft first-pass documentation before it reaches a human, and free the manager to focus on directing the work and resolving the exceptions that come back. The role stops being traffic control and starts being orchestration. This article covers what the manager job has historically absorbed, what moves to an AI agent workforce, the new levers managers actually control, and how performance measurement changes as execution hours compress.
What the manager role has traditionally absorbed
The manager tier sits at the center of every engagement. The role picks up everything that lands between partner-level accountability and staff-level execution:
- Workpaper review
- Staffing puzzles
- Client follow-up
- Review notes
- Issue triage
Look closely at how those tasks split, and two categories emerge. One is judgment work: managers confirm evidence supports conclusions and resolve exceptions before partner-level matters move upward. That work is core to the role and stays with the manager.
The other is coordination work that wraps around the judgment: manually compiling status reports, assembling partner updates from staff emails and half-current trackers, chasing evidence, reconciling versions. Firms built dashboards to reduce it, and dashboards helped, but a large share of the manager's week still goes to it. That coordination layer is where agentic AI creates the biggest opening, and it is the first place the operating model changes.
What moves to Fieldguide's Field Orchestrator
The manager role is moving from coordinator to orchestrator: instead of chasing status and stitching together updates, the manager directs a workforce of AI agents that handle the task-level execution. Those AI agents turn audit procedures into discrete tasks, run them, and return results with documentation drafts and suggested conclusions for the manager to review.
Fieldguide is where that shift becomes concrete for audit and advisory firms. Two pieces work together:
- Field Agents are purpose-built AI agents that execute engagement work under human review.
- Field Orchestrator is the conversational interface managers use to direct Field Agents and see what they've done.
Together, they change what a manager's week actually looks like. Instead of pinging seniors for status, the manager checks Fieldguide's Kanban-style Field Board (a live visual view of engagement state) which shows whether each item is:
- Ready for Testing
- Waiting on Evidence
- Flagged for Review
- Complete
Evidence gets checked at intake, which cuts inbox triage over whether a client upload actually satisfies the request. Partner updates pull from the engagement record instead of a scratch build every week.
The coordination work that used to define the manager's Tuesday now runs in the background, with clear queues and defined escalation paths built into the platform so the workforce stays directable.
The manager's new concrete levers
Once the coordination layer moves to agents, operating control shifts to a smaller set of decisions the manager owns directly. The managers who find these levers early will be the ones partners trust with the model. Three matter most.
Deciding what Field Agents execute
The first lever is scope: which procedures fit Field Agent execution, on which engagements, and when. A high-volume, rules-heavy area like accounts payable (AP) testing is a different call than a judgment-heavy estimate, and Fieldguide makes both easy to configure at the engagement level.
Two features do the heavy lifting. Agent Triggers run Field Agents automatically when workflow events happen, such as a document upload or a completed request, so evidence gets analyzed the moment it arrives instead of waiting for someone to notice it landed. Agent Knowledge grounds every agent run in the firm's own methodology, prior-year work, and standards, so what a Field Agent produces already looks and reads like the firm's work. Together, they give the manager precise control over what runs, when it runs, and what it draws on.
Setting checkpoints and exception thresholds
The second lever is what the manager sees and when. Every Field Agent output is designed for human review, and Fieldguide surfaces the pieces that need judgment first: flagged items, exceptions, anomalies, and matters that belong in front of the partner. Managers can pause an agent run, redirect it when evidence does not match the request, or route the exception upward, all inside the same workspace.
Agent Review Experience adds another layer. Preparers can append context to Field Agent output before it reaches the manager, so the manager reviews prepared work instead of raw output. That mirrors the way review has always worked — a manager reviewing a senior's write-up, not the raw testing — and keeps the accountability chain intact.
Documenting the workflow and the workpaper
The third lever is the file itself. Quality management evidence now extends past signoffs; for agent-executed work, the engagement file may need to retain configuration snapshots and change logs under Statement on Quality Management Standards (SQMS) No. 1, effective December 15, 2025. That is a lot of extra documentation to produce by hand.
Fieldguide produces it in the background. Every Field Agent run generates a Trace of the path from inputs through outputs to reasoning, with citations back to source documents. That gives the manager something concrete to evaluate and gives an inspector something to follow, without adding manual documentation steps to the manager's day. The workflow itself has become a workpaper, and Fieldguide keeps that workpaper current.
How manager performance gets measured
The old scorecard for a manager was realization, utilization, and on-time delivery. Those are still on the sheet, but they were designed for a world where execution hours filled the week. That world is going away.
Execution hours are already dropping. KPMG's FY2024 audit quality report shows average weekly hours worked on audit engagements declined 18% compared with 2020, and firms using Fieldguide are seeing similar compression in individual tasks: UHY reported 20-30% time reductions from AI and automation, with some tasks collapsing from three hours to fifteen minutes.
When execution shrinks, supervision becomes the largest thing a manager does, so that is what gets measured. A modern audit quality scorecard tracks supervision hours as a share of total engagement hours, alongside documentation of professional skepticism and dashboards tied to quality outcomes rather than revenue alone. Managers are graded on checkpoint execution: whether unsupported Field Agent output was caught, missing procedures were surfaced, exceptions were routed correctly, and partner-level matters were escalated.
That is the metric Fieldguide is built to make visible. Every Field Agent run leaves a Trace with citations, every review action is captured in the engagement record, and the manager's judgment shows up in the file where partners and inspectors can see it. Efficiency and checkpoint evidence sit side by side in the same record.
What doesn't change
The accountability chain looks the same as it always has. Public Company Accounting Oversight Board (PCAOB) Auditing Standard (AS) 1201 still places responsibility for engagement performance with the engagement partner, and responsibility for conclusions and assurance still sits with the firm. AI changes how the work gets done, not who is answerable for it. A manager supervising agent-executed work sits inside the same chain, and the Field Agent is treated as a preparer subject to review and escalation.
The standards those responsibilities run through have not changed either. The PCAOB has not adopted anything specific to AI, and its staff update on generative AI addresses the topic under existing standards. Firms are applying a familiar framework to a new kind of preparer.
That is precisely why the manager tier stays central in an agent-led model. It is the control layer the whole engagement runs through, and the managers who build rigor into that layer now are building the version of the role that gets promoted.
Where Fieldguide fits in the orchestration shift
Orchestration works best when firms can manage key parts of the operating layer in one place:
- Engagement work
- Evidence
- Review
- Documentation
Otherwise, teams are stitching together disconnected tools, and the manager's day drifts back to coordination. Fieldguide is the industry's only end-to-end AI-native platform, purpose-built for audit and advisory, with the Agent Workforce, methodology depth, and audit-grade rigor firms need to operate this way. Field Agents execute the engagement work, Field Orchestrator gives managers a conversational way to direct it, and every action lands in the same engagement record so partners and inspectors can see the review path end to end. Book a demo to see how the orchestration workflow runs on an engagement.
Amanda Waldmann
Increasing trust with AI for audit and advisory firms.
