Bot Manager

Table of Contents

  1. Definition
  2. Why it matters for marketing
  3. What a Bot Manager does
    1. 1. Prompt design and logic auditing
    2. 2. Orchestration logic
    3. 3. Performance governance
    4. The 1-to-5 model
  4. Skills and competencies
  5. How to utilize the role
  6. Comparison: Bot Manager vs. related roles
  7. How to measure a Bot Manager
  8. Best practices
  9. Future trends
  10. FAQs
  11. Related Terms
  12. Listen: related episodes of The Agile Brand
  13. Sources

Definition

A Bot Manager is the person responsible for supervising a team of AI agents doing commercial work. Instead of running campaigns or building workflows by hand, a Bot Manager defines what the agents are trying to achieve, sets their limits, audits how they reach decisions, handles the exceptions they escalate, and improves their performance over time.

Greg Kihlström describes the Bot Manager in Marketing Operations 3.0 (De Gruyter Brill) as “the central human role in the hybrid operating model.” The book’s shorthand for the change is a move “from doing to managing.”

See also: Marketing Operations · AI Agent · Human-in-the-Loop (HITL) · Prompt Engineering

Not to be confused with: Bot management in cybersecurity, which means detecting and blocking automated web traffic such as scrapers, credential-stuffing bots, and fake account sign-ups. Akamai also sells a security product called Bot Manager. In marketing operations, a Bot Manager supervises a company’s own AI agents. It has nothing to do with filtering outside bot traffic.

Back to Top

Why it matters for marketing

For years, marketing ops professionals earned their value through platform fluency: configuring automation, building segments, fixing broken syncs, cleaning up data. Those skills still matter. But once AI agents can execute multi-step work on their own, someone has to decide what they should do, check whether they did it well, and step in when they shouldn’t act alone.

Without that person, companies tend to drift toward one of two extremes. Either they keep humans approving every minor step, which wipes out the speed gain, or they hand agents broad authority with nobody watching closely enough to catch drift.

The Bot Manager role is the middle path. Microsoft’s 2025 Work Trend Index makes a similar argument at a broader level. It describes an emerging “agent boss” mindset in which employees build, delegate to, and manage AI agents as part of their everyday work.

Back to Top

What a Bot Manager does

The book breaks the job into three core responsibilities.

Back to Top

1. Prompt design and logic auditing

In Marketing Ops 1.0 and 2.0, auditing meant looking for missing fields or broken integrations. A Bot Manager audits the reasoning behind an agent’s action, not just whether the workflow ran.

The book gives a concrete case. If an agent decides to stop emailing a customer segment, the Bot Manager has to be able to check whether that decision came from real churn signals or from a hallucination caused by bad or biased data. The role also covers prompt architecture: writing the high-level directives and commercial policies (discount limits, brand voice rules, restricted actions) that act as guardrails.

Back to Top

2. Orchestration logic

Agents in marketing, sales, and service need to act from the same customer context without stepping on each other. The Bot Manager designs how they coordinate. That includes what context each agent inherits, when it escalates, and which actions should be suppressed in certain situations. The book’s example is a customer with an open, high-stakes support ticket, who shouldn’t be getting upbeat personalized ads at the same moment.

Back to Top

3. Performance governance

The book argues that agents should be “managed as talent.” In practice that means:

  • Onboarding: testing agent behavior in a controlled environment before granting live authority.
  • Monitoring learning velocity: watching whether an agent’s accuracy improves or drifts as conditions change.
  • Red-teaming: deliberately trying to trip agents up to find logic gaps before customers do.
  • Kill switch readiness: knowing when and how to pause an agent that’s going off course.
Back to Top

The 1-to-5 model

A Bot Manager typically leads a human-agent squad. The book’s “1-to-5 model” pairs one human lead with roughly five specialized agents. The exact ratio isn’t the point. What matters is that the unit stays small enough for the human to exercise real judgment.

Back to Top

Skills and competencies

Chapter 20 of the book lists five capability areas for the role (titled there as CO2 Lead or Agent Orchestrator):

CompetencyWhat it looks like in practice
Reasoning oversight and logic auditingSpotting weak retrieval, missing context, and conclusions that sound plausible but aren’t commercially sound
Value engineering and outcome mappingTying agent work to business outcomes, and noticing when faster isn’t better
Cross-functional orchestrationDesigning how agents share context, authority, and escalation across marketing, sales, and service
Brand stewardship and ethical guardrailsKnowing what the system shouldn’t do, even when it could
Data architecture for machine readabilityEnough data fluency to tell when agents are working from solid information and when they aren’t

The book is clear that this isn’t “a renamed platform administrator.” It’s a managerial role, and the job description should be written around responsibilities, not buzzwords.

Back to Top

How to utilize the role

  • Start with one squad. Assign a Bot Manager to a single outcome, such as onboarding continuity or early churn prevention, with a defined set of agents.
  • Write the job description around five responsibilities: defining operating intent, maintaining orchestration quality, auditing logic and exceptions, protecting guardrails, and linking agent work to commercial results.
  • Give the role a seat in governance. In the book’s model, a Bot Manager liaison sits on the CO2 Council so leadership hears what’s happening in live execution.
  • Upskill from within. Strong marketing ops and automation specialists are natural candidates, with added training in reasoning audit and constraint design.
  • Measure the managed unit, not personal output. Judge the Bot Manager on whether the squad produces better outcomes with stable guardrails.
Back to Top
Marketing Ops Manager (2.0)Marketing Automation SpecialistMLOps EngineerBot Manager
Main focusPlatforms, workflows, reportingBuilding and maintaining automated programsDeploying and monitoring ML modelsSupervising AI agents doing commercial work
Works onCampaigns and the martech stackNurture flows, triggers, scoring rulesModel pipelines and infrastructureAgent goals, guardrails, reasoning, escalation
Key question“Did the workflow run?”“Is the program firing correctly?”“Is the model performing in production?”“Did the agent reason within policy and produce a sound outcome?”
Typical backgroundMarketing technologyAutomation platformsData science and engineeringMarketing ops plus AI governance
Measured onEfficiency, lead flowProgram performanceModel accuracy, uptimeSquad outcomes, escalation quality, guardrail health
Back to Top

How to measure a Bot Manager

The book says hybrid talent should be measured “less by how much work a person touches directly and more by whether the system they govern becomes more effective.” Useful signals include:

  • Outcome quality of the managed squad (conversion, retention, resolution)
  • Escalation quality: are the right cases reaching humans, with enough context?
  • Guardrail health: drift rate, compliance exceptions, error recovery time
  • Hybrid Squad Efficiency Ratio (HSER), a measure of how much productive output one human lead gets from a team of agents
  • Toil, as an anti-metric. Time still spent on manual reconciliation, cleanup, and ritual approvals is a sign the model isn’t working yet.
Back to Top

Best practices

  • Don’t grant agents live authority until they’ve been validated in a test environment.
  • Audit reasoning, not just results. A correct output reached through bad logic will eventually fail.
  • Define escalation triggers in advance (deal size, sentiment, compliance exposure, contract deviation) so the Bot Manager isn’t deciding case by case.
  • Keep the span of control manageable. More agents per person only helps if oversight quality holds.
  • Treat pausing an agent as normal stewardship, not as failure.
  • Run red-team exercises on a schedule, not only after incidents.
Back to Top
  • Formal job titles. Expect Agent Orchestrator, AI Operations Lead, and similar titles to appear more often in marketing and CX job listings.
  • Agent bosses at every level. Microsoft’s Work Trend Index research suggests managing agents will become a general employee skill, not only a specialist role.
  • Guardian agents. Automated auditing agents will take on routine monitoring. The book notes that humans still have to interpret what audits mean and decide what changes.
  • New performance frameworks. HR and marketing leaders will need review criteria for people whose output is mostly agent output.
  • Training programs focused on reasoning audit, constraint design, and escalation handling.
Back to Top

FAQs

What is a Bot Manager in marketing? The person who supervises a group of AI agents doing marketing, sales, or service work. They set goals and limits, audit the agents’ reasoning, handle escalations, and improve performance.

Is a Bot Manager the same as bot management software? No. Bot management in cybersecurity blocks unwanted automated traffic to websites. A marketing Bot Manager oversees the company’s own AI agents.

Who coined the term? Greg Kihlström uses it as the central human role in Marketing Operations 3.0. The book also uses the titles CO2 Lead and Agent Orchestrator.

How many agents does a Bot Manager oversee? The book’s 1-to-5 model suggests about five specialized agents per human lead. The right number depends on how complex and risky the work is.

What skills does a Bot Manager need? Logic auditing, prompt and constraint design, cross-functional orchestration, brand and ethical judgment, and enough data fluency to judge whether agents are working from reliable information.

Is this a new job or a new version of marketing ops? Mostly an evolution of marketing operations. Platform and workflow skills remain, but the center of the job moves from executing work to supervising agents that execute it.

How do you evaluate a Bot Manager’s performance? By the results and reliability of the squad they manage (outcomes, escalation quality, guardrail health), not by how many tasks they personally complete.

What is “logic auditing”? Reviewing the reasoning, context, and evidence behind an agent’s decision, not just the output, to confirm it acted within policy and for sound reasons.

Does a Bot Manager need to code? Not necessarily. The role calls for data and systems fluency and the ability to write clear instructions and constraints. Deep engineering work usually sits with technical teams.

Back to Top
  1. Marketing Operations
  2. AI Agent
  3. Human-in-the-Loop (HITL)
  4. Prompt Engineering
  5. Concept Drift
  6. Machine Learning Operations (MLOps)
  7. Citizen Data Scientist
  8. Learning & Development (L&D)
  9. Human-Agent Squad (1-to-5 Model)
  10. CO2 Council
Back to Top
Back to Top

Sources

Back to Top

Tags: ,

Was this helpful?

Next

Change Control Board