How AI agents change manager workflows

A specific category of knowledge worker has been under-discussed in AI agent content: managers.

Managers don't do the same work as individual contributors. Their day is different. The agents that help them are different. The way they think about agent leverage is different.

This post is about how AI agents change manager workflows specifically, based on what we've heard from manager customers.

The manager's operational reality

Managers spend their time on:

→ One-on-ones with reports → Cross-functional meetings → Team coordination → Performance reviews (drafting + delivering) → Hiring (when actively hiring) → Status reporting upward → Resource allocation decisions → Conflict resolution → Coaching and feedback → Strategic thinking (when there's time)

The pattern: lots of recurring meetings, lots of recurring communications, lots of administrative overhead around people management, and the strategic thinking that's actually the highest-value activity often gets squeezed out.

Managers we've talked to consistently report that the operational overhead has grown over years. Slack pings, meeting load, documentation requirements, performance management bureaucracy. The job has more operational tax than ever.

AI agents could help. The question is which agents help managers specifically.

Agents managers find useful

Sophia (meeting follow-ups).

Manager has 15-30 meetings per week. Follow-ups are crucial for action item tracking. Without agents, follow-ups happen inconsistently.

Sophia attends every meeting via transcript, drafts personalized follow-ups per attendee, sends after each meeting. Action items get assigned and tracked.

Impact for managers: action item completion goes up substantially. Meetings produce results instead of just discussions.

One-on-one prep agent.

Manager has weekly one-on-ones with each report. Good prep takes 5-10 min per person. With 6-8 reports, that's an hour of prep weekly.

Agent reads recent activity per report (Linear tickets, GitHub PRs, Slack messages, recent meetings), drafts prep notes per person highlighting: progress, blockers, concerns to raise, recognition opportunities.

Impact: one-on-ones become more focused. Manager comes prepared. Reports feel seen.

Performance review draft agent.

Annual reviews require synthesizing a year of performance data per report. Manual synthesis is hours per person.

Agent reads year's worth of data per report (project outcomes, peer feedback, accomplishments, areas for development) → drafts initial review covering each performance dimension. Manager edits and personalizes.

Impact: review writing time drops 60-70%. Reviews become more grounded in actual data instead of recency bias.

Status report drafter.

Manager writes weekly/monthly status reports upward. Repetitive structure.

Agent reads team activity (project status, blockers, wins, risks) → drafts status report in the format upward management expects. Manager refines.

Impact: status reporting takes minutes instead of hours.

Hiring helper agent.

When hiring, manager needs to: screen resumes, schedule interviews, debrief panels, coordinate offers.

Agent (combination of Marcus template + custom additions): screens resumes against role criteria, drafts interview invitations, sends scheduling links, compiles panel feedback into structured debriefs.

Impact: hiring throughput up. Time to fill drops.

Team coordination agent.

Manager needs to track who's working on what, identify cross-team dependencies, ensure smooth handoffs.

Agent reads project tools (Linear, Jira, Asana) and team communication, identifies dependencies + risks, surfaces to manager weekly.

Impact: cross-functional friction caught earlier.

What managers should NOT auto-action

Personnel decisions. Promotions, terminations, performance ratings — all human judgment.

One-on-one conversations themselves. Agents prep; managers conduct.

Sensitive feedback. Hard conversations are human work.

Strategic decisions. Hiring plan, headcount allocation, project prioritization — human judgment.

Anything affecting interpersonal trust. Trust is built between humans, not via agents.

The pattern: agents handle the documentation, drafting, and tracking work. The human work of managing — conversations, judgment, trust-building — stays with the manager.

What changes for managers who adopt agents well

We've talked to many managers using Avery NXR. Consistent patterns:

Time recovered for actual management work.

Operational drafting time absorbed by agents → recovered time goes to one-on-ones, mentoring, strategic thinking. The work managers actually want to do.

Better follow-through.

Action items get assigned and tracked. Decisions made in meetings actually happen. The team executes better because nothing falls through cracks.

Better one-on-ones.

Prep agents mean every one-on-one starts grounded in current context. Reports feel their manager pays attention. Trust builds.

More frequent praise.

Recognition agents surface accomplishments managers might otherwise miss. Praise becomes more frequent and specific.

Reduced overhead anxiety.

The "I should send that status report" / "I should schedule that one-on-one" / "I should write that review" background anxiety gets absorbed by agents handling the routine portions.

Sleep gets better.

This is qualitative. Managers we've talked to report less work bleeding into evenings. Operational tax was eating evenings; agents absorbing it returns the evenings.

The manager who tried to automate too much

We've also seen the failure mode.

A manager we worked with built agents for: feedback drafting, performance evaluation drafting, recognition messages, one-on-one summaries, AND auto-sending all of it.

Within a few weeks, his reports noticed. The feedback felt generic. The recognition felt impersonal. The reports' direct experience of being managed got worse, not better.

He pulled back. Kept agents for drafting + analysis. Removed auto-send. Personalized everything before sending.

The lesson: agents can help with manager workflows, but the relationship work has to remain human. Personalizing communications, picking the right words, choosing when to praise — these need human judgment.

The "AI-augmented manager" pattern

The managers we see succeeding have a consistent pattern:

→ They use agents for prep + drafts. Their preparation is faster + better.

→ They personalize every communication. Final delivery is human.

→ They use agents to track + remember. Notes from past one-on-ones, commitments, follow-ups — agents help them remember.

→ They free up time for the human work. Recovered hours go to one-on-ones, mentoring, deeper strategic thinking.

→ They don't pretend the agents aren't there. When reports ask "did AI write this?" the honest answer is "AI drafted, I edited and personalized." Most reports are fine with this.

This pattern produces managers who are more present, more thoughtful, and more strategic. The agent is a tool, not a replacement.

What managers tell us they wish they'd known

"I wish I'd configured Sophia immediately. The meeting follow-up impact alone was worth the whole platform."

"I underestimated how much one-on-one prep matters. My one-on-ones became dramatically better."

"I was too aggressive about automating performance reviews. I had to pull back when reports noticed the generic feel."

"I didn't expect to recover my evenings. That was the biggest unexpected win."

"Trust building with my team improved because I had better context for every interaction."

These are paraphrased but represent real patterns.

What managers considering this should evaluate

→ How much of your week is operational drafting vs. relationship work? → Which operational tasks recur most painfully? → Where do action items fall through cracks? → What management work do you wish you had more time for?

The answers identify which agents fit your specific role. Different managers benefit from different combinations.

The 3-agent starter pack for managers we'd suggest:

→ Sophia (meeting follow-ups) — start here, biggest immediate impact → One-on-one prep agent (custom) — build after Sophia is stable → Status report drafter (custom) — add when you have bandwidth

Resist the temptation to build five agents in week one. Start with three. Add more as bandwidth allows.

The bigger picture

Manager roles have been getting harder for years. More meetings, more documentation requirements, more cross-functional coordination, more performance management bureaucracy.

AI agents are a structural opportunity for managers to recover time for the work that matters: relationships, judgment, strategic thinking.

Managers who figure this out in 2026-2027 will:

→ Manage more people effectively → Build stronger team cultures → Have time for actual strategy → Be present for their teams instead of buried in operational overhead

Managers who don't will continue feeling crushed by the operational tax, even as agents quietly improve productivity for everyone around them.

If you're a manager and AI feels like something to "watch and wait" on — the waiting is increasingly costly. The starter pack is small. The impact compounds.

→ avery.software — Free Desktop tier. The platform built for managers who want their time back for actual management.