Managing Autonomous AI Agents

Show your Agents who is the Boss #

Learning how to use agents is the first step. To succeed in the AI workplace transition, we believe individual contributors need a mindset shift, basically everybody in an organization needs to start thinking like a manager.

Our book Managing Autonomous AI agents teaches people how to undergo that shift, building on accepted practices for transitioning reports into managers, selected and adapted for AI agents. We also cover basic technical information as it relates to the mindset shift.

While the book is about AI agents we wrote the book without using them to create any of the text itself. In the case of one of the authors, the book was written by hand (Pablo likes to write on paper and then types back the changes).

Who Is This For? #

  • Individual Contributors in tech and beyond
  • Professionals anticipating AI integration in their workflows
  • Managers transitioning to hybrid human-AI teams
  • Anyone who wants to hedge against skill obsolescence and lead the AI transition

What You’ll Learn #

  • The Manager of AI Agents Transition: How to shift from task-oriented execution to outcome-driven leadership, commanding a team of AI agents as your direct reports.
  • The Three Pillars of Success:
    1. Mindset: Cultivating growth, outcome-driven, and collaborative thinking.
    2. Management: Applying scientific management and modern leadership to AI workflows.
    3. Technical Operations: Mastering evaluation, security, privacy, and failure modes.
  • Hype vs. Reality: A critical look at what AI can do today, what’s overpromised, and how to navigate procurement, shadow AI, and accountability gaps.
  • Real-World Frameworks: The 7-14-30 day onboarding plan, centaur vs. reverse-centaur models, and domain-specific use cases.

Three Pillars #

We focus the transition on three foundational pillars.

1. Mindset Shift #

Mindsets are the operating system of the brain. What you do when you not put too much thought on it. Unconscious beliefs that shape our actions. The AI transition requires moving from:

  • Fixed to Growth: Embracing “I’m not good at AI yet” instead of “AI is too complex.”
  • Task-Oriented to Outcome-Driven: Focusing on the Pareto principle—what truly moves the needle?
  • Individual to Collaborative: Developing a “Theory of Mind” for AI, treating agents as co-creators, not just tools.

2. Management Skills #

Managing machines is different from managing humans. This book bridges classical and modern management:

  • Scientific Management (Taylorism): Optimizing processes, measuring time-motion, and standardizing working conditions for AI.
  • Directive & Collaborative Styles: Knowing when to command and when to empower AI agents.
  • Clear Division of Labor: Humans handle planning, supervision, and training; agents handle execution.

3. Technical & Operational Skills #

Autonomous agents require robust infrastructure. You’ll learn to:

  • Evaluate Rigorously: Move beyond “it works” to systematic benchmarking and LLM-judge validation.
  • Navigate Security & Privacy: Understand data exfiltration, data consent, and the risks of shadow AI.
  • Handle Failure Modes: Address hallucinations, context management, and operational reliability.

The Authors #

This book started as Saturday morning breakfast conversation about the future of data science. What is left for people to do in the field? And the conversation converged into a book proposal, a year of book writing and the book itself. (Yes, the authors are married, which on itself makes the book unusual.)

Our experience with technology and takes for the future are very different. In Managing Autonomous AI Agents you will thus be exposed to blanket optimism peppered with technological skepticism. Instead of blending our voices into a neutral tone, we let them shine through and expect our readers to decide based on their own experiences and inclinations.

Annie Ying #

Annie is the good cop. With 20+ years of experience in data science, AI and software engineering. Some of the career highlights include her leadership in shipping a critical feature to enable self-hosted models for GitLab Duo Agent Platform, delivering a company priority #1 data science project at Cisco, and award-winning contributions to suicide prevention through NLP cited by WHO.

Annie holds a PhD in AI in Software Engineering from McGill University and is a author of eight patents and over 20 peer-reviewed publications, two of which received distinguished paper awards from ACM. She is currently an Engineering Manager at GitLab.

With eight years of management experience spanning corporate and startup settings, she is passionate about empowering individual contributors to equip themselves with the management mindset and skills needed to effectively operate in the new world of AI agents.

Pablo Duboue #

Pablo is the grumpy cop. He is passionate about improving society through technology. With 25+ years programming and building AI systems, he holds a doctoral dissertation in machine learning for text generation at Columbia University. He was part of the original IBM Jeopardy! Watson team. He is currently a core infrastructure engineer at the Internet Archive.

Dr. Duboue has worked in Research Centers, Universities, Companies, and Startups. He has spent the last 15 years helping get innovation into the market by consulting for AI startups. Working closely with co-founders and technologists, he has seen what worked (and what did not) and help share that experience across the industry.

Other career highlights include authoring The Art of Feature Engineering (Cambridge 2019), teaching in three different countries, joint research with more than fifty co-authors, a best paper award in the Canadian AI conference industrial track, and to have consulted for a start-up acquired by Intel Corp.