A team member has just created an AI agent workflow.  It features nodes, triggers, decision branches and fallback paths, all looking impressive on a canvas.  Now, ask them what they called the diagram.  

In short - It’s the agent canvas in Microsoft Copilot Studio and the workflow builder in Azure AI Foundry.  These are different tools, teams and decades of technical experience, yet the diagram is the same.  

This experience designer’s twenty-year-old drawing has been rebuilt from scratch to map AI agent workflows because the need for it has never gone away. 

That’s no coincidence – it’s proof.

The Diagram Got Rebuilt. In Every Room That Needed It.

This is the part worth sitting with.

Engineers or solution architects building AI agent workflows didn't invent a new kind of diagram. They recreated the experience map — independently, out of necessity — and gave it a new name. 

Because when you are trying to model what happens to a process at every decision point, every handoff, every failure state, every moment a human might need to intervene — there is only one shape the thinking takes. There has only ever been one shape.

Open Azure AI Foundry's workflow builder. Every node is a decision point. Every connection is a handoff. Every fallback path is an edge case someone had to anticipate, consciously, before the system went live. Each branch represents a possible user state. Each conditional is an assumption about what the user knows, expects, or does next. The people building these flows are making the same decisions journey mappers make — they are just making them inside a development tool rather than a workshop.

Open Microsoft Copilot Studio's conversation canvas. User intent, misunderstanding, escalation, recovery — all modelled, all branching, all mapped before the system touches a single real user. 

Open Zapier's agentic flows. Exactly where a process branches, stalls, or requires a human to step back in — visible, connected, deliberately designed.

Show any of these canvases to someone who has spent years doing service blueprinting or experience mapping. Their expression is not confusion. It is recognition. 

Because the boxes changed shape and the arrows changed colour, but the underlying logic is identical: model the human path, surface the decision points, design for what happens when things go wrong.

I remember looking at the updated version of that same laptop service chatbot — rebuilt a year later inside Microsoft Copilot Studio. The team walked me through their conversation canvas. Topic nodes for request type. Branch conditions for device category. A fallback path for unrecognised inputs. An escalation trigger when the request sat unacknowledged beyond a set time. They'd spent two weeks building it. When I asked them what they'd call the diagram, they said workflow. I said: this is a journey map. 

You've got touchpoints, decision branches, failure states, and a recovery path. You built an experience map. They looked at it again. Then one of the engineers said — quietly — that's exactly what it is, isn't it.

The diagram was never obsolete. It was being redrawn in a different room, by people who hadn't been told it already had a name, because the work kept requiring it.

That is not a coincidence. That is what happens when something is fundamentally necessary.

Why the Same Diagram Keeps Getting Redrawn

You know why the experience map keeps showing up in different forms across various tech tools and decades? It’s not just a throwback to the past or a habit. 

It’s because the diagram is based on three core principles that remain constant, no matter the technology.

Critical thinking — the ability to evaluate what the system produces against human reality.

An AI workflow can pass every internal check and still confuse the people using it. The logic is clean. The outputs look correct. 

And somewhere, a real person is making the wrong decision because what the system did and what they thought it did are two completely different things. 

Someone has to look at that flow before it goes live and ask the question the tool never asks: does this actually match how a person thinks when they're under pressure, distracted, or just trying to get their job done? That ability doesn't come from a better prompt. It comes from years of watching real people use real systems — and developing a feel for where things quietly go wrong before anyone says anything.

Emotional intelligence — the ability to work with the humans the diagram is about.

The most useful experience maps were never just documentation. 

They were the thing that got a product manager, a business analyst, an IT lead, and a department head into the same conversation — and forced them to agree on something before it became a problem in production. Reading the room. Knowing which concern isn't being said out loud. 

Translating between people who think in data and people who think in outcomes. That is what makes the map useful beyond the meeting it was built in. An AI tool can generate the diagram. It cannot sit in that room and sense what needs to shift before everyone leaves.

Adaptability — the ability to carry the question across every surface it lands on.

From websites to mobile apps. From mobile apps to cloud platforms. From cloud platforms to AI workflows. From AI workflows to whatever comes after them. The people who stayed genuinely curious about what a user experiences at each step didn't have to reinvent themselves every time the technology changed. 

They just moved the question to a new surface. The question is always the same: what does the person using this actually understand right now? The surface keeps changing. The question doesn't. And because the question travelled, the diagram kept getting redrawn wherever the work required it.

The laptop service chatbot is a small example. But I've seen the same pattern in CRM deployments, in procurement approvals, in HR onboarding flows, and now in AI agent rollouts. The technology in the room changes every few years. The question that gets skipped stays the same: what does the person using this actually understand at each step? Every time that question gets answered late, the cost is the same — not a technical failure, but a trust failure. A system that works correctly and feels wrong to the people inside it.

These three things are why the diagram kept reappearing. Not because experience mappers advocated for it. Because every team that tried to work without it eventually needed it — and rebuilt it from whatever tools they had in the room.

The Map Your Team Is Already Drawing. Without Knowing It.

Most enterprise teams building AI workflows right now are doing experience mapping. They just haven't named it that. And because they haven't named it, they're doing it incompletely — without the strategic layer that makes the diagram more than a technical artefact.

Three places where the map is being drawn without being recognised:

Every agent decision point is a human moment in disguise. When an AI workflow branches — when it escalates, stalls, or hands off — something happens to the person on the other side of it. They either understand what just occurred, or they don't. Teams that treat these as purely technical choices build systems that work in demos and confuse people in practice. Teams that ask the human question first — what does the user expect here, what happens if this goes wrong, what do they do when the system doesn't behave as they assumed — build something people can actually trust.

The scenarios still have to be mapped. Every single one. This is the part that hasn't changed at all. Before any AI workflow goes live, someone needs to walk through the best case — everything works, the user gets what they needed. Then the failure case — the agent misunderstands, acts on incomplete information, or does something the user didn't intend. Then the edge cases — the user who does something unexpected, the data that doesn't fit the pattern, the moment the system has to make a call it wasn't designed for. Those three scenarios — positive, negative, edge case — are where the real decisions get made. 

Not in the code. Not in the system prompt. In the thinking that happens before either of those exist. That thinking is scenario mapping. And skipping it is precisely where most enterprise AI rollouts start accumulating the problems they'll spend the next six months explaining..

The map is already being drawn. The teams that know what they're drawing will build something people can actually use. The teams that don't will build something technically impressive that quietly fails the humans inside it.

The Diagram Never Left. Neither Did the People Who Know How to Read It.

Every technology wave declared the basics optional. Every technology wave ended with a team drawing the map they should have drawn at the start — reconstructing the user's path, surfacing the assumptions, identifying where the designed experience and the human experience had quietly stopped being the same thing.

The diagram kept getting redrawn. By engineers. By automation architects. By product teams who might have never attended a journey mapping workshop in their lives — and who produced journey maps regardless, because the work required it.

The need didn't expire. The need never expired. It just migrated — from web flows to mobile journeys, from service blueprints to conversation canvases, from swimlane diagrams to node-based agent workflows — because wherever there are humans moving through systems, someone has to map what they understand at each step.

That has always been true.

And the people who spent years learning to do it — who can walk into a room, look at an agent canvas or an AI workflow diagram, and immediately see the human experience that the technical architecture forgot to include?

They are not redundant.

The basics stayed intact. The diagram kept getting redrawn.

The people who mastered it are now the most valuable in the room.

P.S. The experience map didn't just survive the AI era. It got rebuilt inside every agent canvas, node editor, and workflow tool — by people who needed it badly enough to reinvent it from scratch, without knowing it already had a name.