Image: GPT-4

TL;DR: Traditional journey maps fail to capture the complexity of agentic AI experiences. Designers must evolve from creating static deliverables to designing adaptive systems, focusing on simulation over documentation, cognitive load, symmetry breaking points, experience debt, graceful degradation, and ethical boundaries as explicit design elements.

I recently had a moment every designer dreads: staring at my carefully crafted - and unfinished - journey map for a new conversational AI assistant, and realizing it was fundamentally inadequate. I hit a wall because I realized that the AI agent’s behavior wouldn’t follow the neat, predictable paths I’d drawn. The agent would learn, adapt, and make decisions autonomously in ways traditional journey maps simply couldn’t capture. How might I rethink this tried-and-true design artifact for a world where experiences aren’t fixed pathways but evolving conversations between humans and intelligent systems? How might I do a better job of communicating the unpredictability of a non-deterministic technology?

As Edmundo Ortega from Machine + Partners said:
Adoption of technology means adaptation. To win, you’ll need to change the way you work.

In the age of agentic AI, this has never been truer for designers and design leaders. So to be specific, here’s how I believe the thinking around journey mapping for agentic AI experiences could evolve:

1. From Documentation to Simulation

Traditional journey maps document what happens. Ideally, tomorrow’s journey maps could “simulate” what might happen through dynamic models that can run scenarios across different AI delegation patterns and predict emergent behaviors before they occur.

2. From Emotions to Cognitive Load

We’ve always tracked emotional highs and lows, but now we need to map the invisible cognitive burden. When does managing AI agents become more taxing than doing the task yourself? Where are the hidden costs of delegation and oversight?

3. AI Symmetry Breaking

Journey maps must now identify critical “symmetry breaking points” - where seemingly minor differences in user inputs cause AI systems to follow dramatically divergent paths, creating entirely different experiences despite similar starting contexts. These divergent paths will be difficult to predict now that AI has the “agency” to learn from interactions, reprioritize goals, and make contextual decisions that weren’t explicitly programmed.

4. Mapping Invisible Experience Debt

Each AI interaction creates “experience debt” - small inconsistencies or errors that compound over time. Our journey maps need to visualize this accumulation and identify key moments where this debt must be “paid down” through human intervention.

5. Designing for Graceful Degradation

Rather than assuming consistent functionality, we need to map how experiences gracefully degrade when AI systems encounter edge cases or limited connectivity. The best designs now fail well - not catastrophically.

6. Ethical Boundaries as Design Elements

The guardrails we place on AI systems are becoming explicit design elements. Journey maps should visualize not just what AI agents can do, but what they explicitly cannot do - making ethical boundaries visible rather than hidden constraints.

This evolution represents a fundamental shift for designers. We’re moving from creating static deliverables to facilitating dynamic systems. Our value isn’t in producing polished documentation but in communicating the underlying principles of experiences that continuously evolve.

The designer’s role is transforming from architect to gardener—we’re no longer just building structures but cultivating environments where experiences grow and adapt.

Instead of defining every interaction as a step in a journey map, we’re describing the conditions and boundaries that enable productive human-AI collaboration while allowing for personalization, complex unexpected behaviors, and interactions that may arise from agentic AI systems.

This experience had me questioning whether I was even thinking about it the right way: was a journey map needed for this use case? How can I leave space for when things go sideways…e.g. unpredictable or biased responses, randomness, and hallucinations? For this reason, I’ve come to believe it’s necessary for designers to expand beyond a “traditional” approach to creating deliverables and embrace other cross-disciplinary communication methods.

A designer’s educational and communicative role remains vital, but now requires us to translate complex adaptive systems into understandable frameworks that users and stakeholders can navigate with confidence.

#DesignFutures #AgenticAI #ServiceDesign #ExperienceDesign #DesignInnovation #humancentereddesign #journeymapping