Remote Frog

Constellations is a coaching and systemic therapy technique. The idea is simple: you place the people around you in a given system (your family, your team, your company) — and you draw connections among them in a chart. That’d be, properly said, a graph. The distance and position between them reveal dynamics you can’t see when you’re just thinking abstractly.

What is a professional constellation?

Well, it’s a constellation you compose to understand your relationships at work. And, more importantly, to intentionally act on them. Imagine your work environment as a system with several actors:

• People with power over your career — those who decide promotions, projects, or your continuity
• Allies — those who support you, give you visibility, or have your back
• Stakeholders — those who depend on your work or whom you need to move forward
• Tensions — relationships with friction, competition, or latent conflict
• You — at the center of the system, connected to everyone else

A professional constellation places all these actors on a visual map where:

• Vertical position reflects hierarchy or influence
• Horizontal distance reflects relational closeness or distance
• Connection thickness reflects current relationship intensity
• Line style (solid or dashed) reflects whether there’s harmony or tension

It’s the kind of diagram that, once you see it, makes you think: ah, that’s why I feel this way at work.

The problem is, you sometimes need someone with all the perspective of your interactions, full access to your communications, and the understanding of what’s happening at work, but that is not you. Not you or you’ll put in the graph what you already know!.

And… why don’t we ask an AI agent to read all our communication exchanges; your emails, your participation in project management tools, Slack…? And after that to draw your work system constellation?

My prompt

Excalidraw works perfectly for these maps. It’s simple, easy to use, and has a simple but cool look. So I use it a lot. I use it, for instance, to draw parts of the ETL pipeline I work with. But it can also be used to draw relationships:

Excalidraw internal format is simply JSON — a format any AI can generate. So I asked an Agent to read my communications in Linear, email, Slack, P2s, etc., and to take a look at my company’s org chart so that it could draw up my professional constellation. This was the literal request:

I want you to create a diagram with the most important actors I need to build relationships with after studying my communications. The actors should be distinguished by three levels of importance (CEO on top), maybe by the box format. My box is in the center. The relationships (arrows) can be of three types: thicker or thinner depending on my current activity with them, and solid or dashed if there have been conflicts.

The AI returned a JSON that I saved as map.excalidraw and opened directly in Excalidraw. It’s not perfect — some boxes overlap, some arrows cross where they shouldn’t — but it gives a very clear starting point to work from. All I need to do is to iterate on the agent definition -not much-. And Excalidraw is precisely the tool I’ve always used for this kind of mental map.

What’s the result and how the JSON look like

The result was:

And these are a couple of fragments of the generated JSON. First, a box representing a person:

{
  "id": "box_ceo",
  "type": "rectangle",
  "x": 553,
  "y": 20,
  "width": 175,
  "height": 65,
  "strokeColor": "#e67700",
  "backgroundColor": "#ffe8cc",
  "fillStyle": "solid",
  "strokeWidth": 3,
  "roundness": { "type": 3 }
}

And its text label:

{
  "id": "text_ceo",
  "type": "text",
  "x": 558,
  "y": 30,
  "text": "CEO\nTop of the org",
  "fontSize": 13,
  "fontFamily": 2,
  "textAlign": "center"
}

Once you see the structure, it’s obvious: the AI agent can interpret what’s happening around you, and it understands the JSON schema; it can generate entire diagrams. The canvas is just JSON.

Why AI-built Constellations work?

Systemic constellations work because they externalize what’s in your head. While relationships live only as diffuse feelings — “I think my boss doesn’t see me”, “that colleague makes me tense” — you can’t work with them. When you put them on a map, they become concrete.

The diagram forces you to answer questions you normally avoid:

• Who actually has power over my career?
• With whom do I have strong relationships, and with whom have I neglected them?
• Where is there tension I haven’t resolved?
• Who should be closer to me and isn’t?

And once you see it, you can act. Send that message. Request that meeting. Resolve that friction.

However. When you draw your own constellation, you bring your preconceptions and prejudices, your vision of reality, and you risk answering those questions, to some extent, in terms of ‘what you already know’ rather than simply analyzing what’s happening from the outside. Your constellations are subjective. It’s human.

Asking an AI agent makes it possible to draw an objective constellation that is based upon facts -communication + the org chart in the example. But you can also:

• Enrich the input data with extra facts (something important you know and can describe to the agent)
• Restrict the data analysis to a given period in time: last 3 months of communications, last 6 months… to more accurately represent the current scenario
• Make constellations to evolve dynamically by redrawing it once a week, for instance, to have it always present and always updated.
• Ask the agent to consider the facts differently, for instance giving more weight to things happening over P2 than over Slack…
• Factor in the way in which people usually talk. In a multicultural community, we know some people’s messages look extremely kind even when they are killing you, while others are apparently blunt and sharp even to praise others.
• And… Iterate on a built Constellation if we think something doesn’t reflect reality. We can ask why of a given relationship or conflict (it’s very good at unearthing conflicts!!) or to factor in something you’ve or the agent have skipped.

It’s like having an illustrator who understands natural language instructions and delivers a working file.

A personal note

I’ve done this exercise every year or so in my last two jobs, like for the last 20-25 years. It’s a powerful tool that forces you to reflect on your relationships and to answer some questions you usually don’t have time -or the right state of mind- for. But I found two problems:

• As I said, you organize your thoughts and beliefs, but you usually don’t discover things you didn’t know already.
• The map takes too much time to make every couple of weeks. But it starts becoming obsolete as you draw it. One month later, the map is no longer so accurate. I made some amendments -I loved Excalidraw precisely because modifications are sooo easy. But at some point, the graph becomes no longer useful.

AI agents cover these two blind spots. On one hand, they unearth relationships or look at them from a different perspective that makes you think and enriches your understanding of the scenario. It forces you to see things you knew but hadn’t articulated: relationships you had oversaught, people with influence you barely knew, tensions you had been dragging along without resolving.

And you can also ask for a new map to be generated every week if needed, and compare changes with the previous chats -very useful conversation too-.

The diagram doesn’t cover anything on its own. But it gives you a hint of some things you’ve skipped, drawing them on a map, so you know which way to walk.