AI Agent Squads in Project Management

A project manager is like a circus performer spinning two plates in each hand, another on their nose, while hopping on one foot and answering questions about project status, budget, scope, and team assignments. LLM-based agents free the delivery manager from repetitive tasks, allowing them to focus on delivering value to the client. They don’t replace human judgment, but they do create space to make a real difference.

These agents are particularly useful in structured, repetitive workflows. It’s not about having an “all-powerful agent,” but about applying the principle of divide and conquer: using multiple specialized agents, coordinated by a central one acting like a conductor.

Each specialized agent handles specific, well-defined tasks and improves over time, while a central coordinator routes requests, manages changes, and maintains context. It’s all backed by models that understand complex intentions and intelligent filters that keep conversations clean (filtering out profanity, violence, aggression, etc.).

Agents share a centralized context that stores intentions, decisions, and history, enabling coherent, natural conversations. Each agent is a modular, scalable, and swappable component. Here’s an example of an interactive flow showing how requests are handled, how conversation shifts are managed, and how context is maintained:

1. Conversation Start

The user writes a prompt like: “Can you give me a summary of the ACME project status?”

2. Enters the Orchestrator

The message reaches the orchestrator agent, which:

  • Detects the user’s intent
  • Identifies key data like project names
  • Uses conversation history to maintain context

3. Assigns the Right Agent

It passes the baton to the reporting agent, specialized in generating reports. This agent accesses the data, checks for deviations, and prepares a summary.

4. A Change of Direction (Conversation Shift)

The user follows up with:
“Okay, and what risks did we identify this week?”
The orchestrator detects the new intent, activates the relevant risk agent, and transfers the context, no need for the user to repeat themselves. A clear upgrade over traditional call center interactions where users often have to start from scratch.

5. Smooth Transition

When the orchestrator connects to the risk agent, the user doesn’t have to repeat anything. Context is already stored. The result? The conversation flows naturally, maybe even better than natural.

6. Outcome

The risk agent carries out the task and responds to the user, confirming the action. The orchestrator then handles any new requests or redirects as needed.

To trust the system, it must be auditable: logs, model versions, traceability, and minimal explanations for each decision, with the option to escalate to a human if something goes wrong. The key is clarity: knowing what each agent does and why, to ensure transparency and control. Identifying the user’s intent and preserving context is what creates a coherent experience.

In Michael Ende’s novel Momo, a little orphan girl becomes the guardian of people’s time. The “grey men” steal time under the pretense of efficiency, pushing people into a world of rush and anxiety. Momo, with her kindness and ability to truly listen, brings back time and joy to the city, restoring humanity.

In real life, managers also face their own “grey men”: time-sucking tasks and tools that create more work than they save. Agent squads aren’t here to replace intuition or leadership, they’re here to free up quality time.

Momo knew that time isn’t something you hoard like gold, it’s something to be honored. We need to listen better, restore meaning to every minute. These agents can help reclaim stolen time, to think, to breathe, to look around, and to fight the “grey men.”

Jorge Alarcón
+ posts

Scrum Master at Keepler. “Working with people to build products that solve problems. Digital transformation is the new industrial revolution based on the fractal creation of team-based development systems. I collaborate with companies to understand problems and develop solution strategies.”

2 Comments

  1. Álvaro López García

    Really interesting, Jorge. Seeing how this develops within Keepler will be very positive. Congratulations!

    Reply
  2. Andrés García Dominguez

    It is pretty impressive, i am keen on hearing more about this topic.

    Reply

Leave a Reply

You May Also Like

The AI Orchestra: Who’s Who in the New Era of Management

The AI Orchestra: Who’s Who in the New Era of Management

Artificial Intelligence is not just a technological revolution; it is an organizational revolution. Many companies make the mistake of buying the technology before building the structure to support it. The result is well known: "pilots" that never scale, investments...

read more
Top 3 AI Use Cases with More Potential Impact by Industry

Top 3 AI Use Cases with More Potential Impact by Industry

Measuring the impact of AI in our use cases is crucial for demonstrating the value of our solutions and ensuring alignment with your business objectives. By tracking key performance indicators (KPIs) such as efficiency gains, cost reductions, and revenue increases, we...

read more

Discover more from Keepler | The AI Enabler Partner

Subscribe now to keep reading and get access to the full archive.

Continue reading

Privacy Overview

This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.