The "agent ops" role: a new function emerging in companies

A specific role is emerging at companies that have meaningfully adopted AI agents. We don't have a universal title for it yet. Different companies call it different things:

→ "Agent Operations Lead" → "AI Operations Manager" → "Agent Engineer" → "Workflow Automation Specialist" → "Internal AI Lead"

The role does similar work across companies. Here's what we've seen.

What agent ops actually does

The role focuses on:

→ Building and maintaining agents. Configuring templates, refining prompts, tuning thresholds, building custom agents.

→ Coaching teammates on agent use. Training new users, troubleshooting issues, sharing best practices.

→ Monitoring agent health. Reviewing audit ledgers, checking output quality, identifying drift, retiring unused agents.

→ Connector management. Setting up new integrations, handling auth refresh, troubleshooting broken connections.

→ Compliance and audit. Ensuring agents meet company policies, maintaining audit trails, preparing for audits.

→ Strategic agent portfolio. Identifying what should be automated next, prioritizing, retiring agents past their value period.

→ Cross-functional coordination. Working with HR, finance, marketing, support, etc. to identify high-value agent opportunities.

It's somewhere between a sysadmin, a workflow analyst, and an internal consultant. Comparable to "marketing ops" or "sales ops" in earlier eras.

When companies need this role

We've watched companies adopt agent ops at different stages:

Companies <50 employees: Usually don't need a dedicated role. Champions in different functions handle their own agents. Founder or COO oversees.

Companies 50-150 employees: Often have a part-time agent ops function. Sometimes a product ops or RevOps person also owns agent operations. Sometimes a senior engineer with capacity.

Companies 150-500 employees: Often a dedicated agent ops role emerging. Reports to COO, CTO, or VP of Operations.

Companies 500+ employees: Often a small team. 2-5 people dedicated to agent operations across the company. May be embedded with the IT/operations function.

The pattern: as companies scale, the volume + complexity of agent operations grows past what champions in each function can handle on top of their day job.

What we hear from people in agent ops roles

We've talked to many people who do this work, even if their titles differ. Common patterns:

They come from varied backgrounds. Some are former engineers. Some are former marketing/sales/HR ops. Some are former IT. The role is new enough that there's no canonical pedigree.

They're embedded in their company's workflow. They understand business processes deeply. They translate process knowledge into agent configurations.

They balance technical + business work. Configuring agents requires technical comfort. Identifying what to automate requires business judgment. Both are core to the role.

They become company-wide multipliers. A good agent ops person can affect productivity across many functions simultaneously.

They face an interesting career question. The role is new. Where does it go next? Some pivot to ops leadership. Some pivot to product. Some become consultants.

What makes someone good at this role

Traits we see in successful agent ops folks:

→ Operational mindset. Comfortable with process thinking, repeatable workflows, instrumentation. → Technical comfort. Can read YAML configs, understands API basics, debugs problems systematically. → Business savvy. Understands the workflows of multiple functions enough to identify automation opportunities. → Strong communication. Can explain agent behavior to non-technical stakeholders. Can train users effectively. → Patience. Agents need iteration. Quality comes from multiple cycles of refinement. → Strategic thinking. Knows when to build, when to retire, when to scale.

This is a wider skill profile than most existing roles. People who fit it well are valuable.

What companies should pay agent ops

Compensation data we've seen:

→ Part-time agent ops (~20-50% of someone's role): added to existing role, often without comp bump unless role grows → Full-time agent ops at small companies (~100-150 employees): $80-130K depending on geography → Full-time agent ops at mid companies (~150-500): $100-160K → Agent ops at larger companies: Lead $130-200K. Multiple ICs $90-140K.

These numbers are rough. The role is new enough that comp benchmarks are still forming.

The math: if an agent ops person enables 10-30 hours/week of productivity gain across the company, their salary is small compared to the value created.

What companies often get wrong with this role

Mistake 1: Assigning to someone with no bandwidth.

The role requires actual capacity. Adding agent ops as 5% of someone's existing role usually means it doesn't happen. The agents languish.

Mistake 2: Hiring a senior engineer for this.

Some companies hire senior engineers thinking the role needs deep technical skills. Engineers often bristle at the role's operational + business focus. The right hire is someone who LIKES this work, not someone who tolerates it.

Mistake 3: Underestimating the cross-functional component.

Successful agent ops requires building relationships with leaders in HR, marketing, sales, support, etc. It's not just configuration work. It's also internal consulting.

Mistake 4: Not giving the role enough authority.

Agent ops needs to make decisions about retirements, modifications, priorities. If they need to escalate every decision, the role doesn't work.

Mistake 5: Measuring on outputs only, not outcomes.

Number of agents created is a vanity metric. Hours saved, customer impact, employee satisfaction with their automated workflows — these are real outcomes.

What we recommend for companies considering this role

If you're at the stage where you're wondering about agent ops:

→ Start with one person doing it as 40-50% of their role. Validates need before full-time hire.

→ Pick someone who LIKES this work. Not someone available. Someone enthusiastic.

→ Give them authority to make decisions. Agent retirements, priority calls, configuration choices.

→ Measure outcomes, not outputs. Time saved across the company. Employee satisfaction. Customer impact.

→ Connect them with peer practitioners. The community is small but growing. Help them build network.

→ Plan for the role to grow. Successful agent ops often expands. Build career path proactively.

The peer community for this role

A small but growing community exists for people in agent ops roles. Mostly informal:

→ Slack communities specific to specific platforms (Avery NXR has one) → Twitter/X discussions among practitioners → Conference talks at events like RecOps Summit, AI conferences → LinkedIn groups slowly forming

The community is still finding itself. People doing this work would benefit from connecting with others doing similar work.

If you're in this role or considering moving into it, finding peers is valuable. The work is new enough that peer learning beats abstract advice.

What this looks like in 5 years

We think this role will become standard in companies of any meaningful size by 2030:

→ Standard title (probably "AI Operations" or "Agent Operations") → Established career path → Clear comp benchmarks → Professional certifications → Conferences and communities → Books and courses

Companies that have this role established by 2028 will operate at structural advantage over companies that don't.

It's the kind of role transition that happened with: → DevOps in the 2010s → Customer Success in the 2010s → RevOps in the 2010s-2020s → Marketing Ops in the 2000s-2010s

Each became essential for scaling. AI Ops will follow the same trajectory.

What we're doing to support this role

For Avery NXR customers specifically:

→ Documentation aimed at agent ops practitioners → Templates for common workflows that practitioners can customize → Audit ledger features designed for ops use cases (not just compliance) → Connector management features for handling auth refresh at scale → Community for Avery NXR power users

We're investing in this role because it's the role that makes platforms like ours actually deliver durable value at organizational scale.

What this means for individuals

If you're considering this role:

You're early. The role is new. There's career risk in betting on a not-yet-standard function.

You're also early in a good way. Roles that become standard reward early movers. The people who became RevOps leaders in 2018 had outsized careers.

Build the skills now. Agent configuration, prompt engineering, workflow design, cross-functional collaboration, audit/governance. All valuable regardless of whether your specific title becomes standard.

Build the network now. Connect with others doing similar work. The community will grow into formal infrastructure over the next few years.

→ avery.software — Free Desktop tier. The platform that AI ops practitioners can demonstrate value with quickly.