Scenario-Driven Design

SAP performance scenario script preview with scenes and details

User roles, JTBDs, and user flows are hallmarks of user-driven design, but novel AI experiences can both benefit from and empower detail-rich technical screenplays, making providing a consistent basis for designing screens, scripting agent interactions, and testing whether the experience holds together across non-deterministic scenarios.

Role
Senior UX Designer, partnering with product, research, and engineering
Timeline
Feb 2025 – Present
Outcome
The scenarios became a backbone of the product team's process, and the experiences shipped, were adopted by users, and appeared at major industry conferences.

Context

Scenario flow for a configuration drift remediation: the script content on the left is broken into beats, and each beat maps to screens in a UX flow across the workload dashboard and Copilot chat.
Early explorations leveraged narrative flows but lacked the technical details that really impact a solid experience design.

When we began designing for AI-based chat experiences, we quickly realized prescriptive action-by-action user flows didn’t provide the right details for envisioning and developing these novel experiences. the team was introducing AI agents to help validate system conifguration and troublehsoot workload issues.

SAP performance scenario script mapped to chat UI: the user prompt, the agent investigation trace, and the agent response sections of the script are color-coded and shown beside the matching parts of the chat interface.
Our scenarios grew in technical details as we iterated, but UX was largely dependant on SMEs on the product side.

The design challenge was not simply adding chat to a dashboard. We needed to show how the agent’s investigation related to the user’s context and intent, how it surfaced evidence, what that evidence was, and what users would see as they moved through changing contexts towards an actionable conclusion.

A screen showing a configuration dashboard with a contextual chat window
A script should provide all the details needed to test and ensure progressive disclosure and continuity.

The problem

Generic user flows can look convincing screen-by-screen and still fall apart as a continuous whole. Plausible but generic or isolated numbers in a dashboard may prove the utility of a component, but do not establish a real world narrative where a user could recognize a problem, follow the evidence, understand the agent’s conclusion, and decide what to do next. We needed realistic scenarios with enough technical detail to test that journey, not just confirm we have all the right patterns in the UI.

A diagram showing a prescriptive UI-centric user flow.
A user flow prescribing what the user does in the UI for each step.

The method: technical screenplays

My approach builds on scenario-based design but pushes it further. Each scenario is written like a technical screenplay, specific enough that every screen and every AI response has to answer to it and can demonstrate a synthetic but realistic storyline.

User flow: What the user does with the UI

two diagrams comparing user flows with screenplay flows
clicks, selects, edits, saves, navigates

Scripted flow: What the user uses the UI to do

two diagrams comparing user flows with screenplay flows
investigate, decide, configure, remediate, verify

What a screenplay contains

A visual showing how the script details map to elements of a screenplay.
The elements of a script.

How scenarios run through the designs

SAP performance scenario script mapped to chat UI: the user prompt, the agent investigation trace, and the agent response sections of the script are color-coded and shown beside the matching parts of the chat interface.
How the script maps to AI chat components and patterns.

The screenplay supplies consistent data for mockups, connects the screens into a user journey, scripts agent interactions, and drives prototypes. It follows the consequential steps from the user’s initial need through investigation to a supported outcome—whether that is a fix, a diagnosis, or a clear next action

Scenario flow for a configuration drift remediation: the script content on the left is broken into beats, and each beat maps to screens in a UX flow across the workload dashboard and Copilot chat.
An example showing how I map a scripted story across an end-to-end experience.

One scenario, two surfaces

This video shows a prototype built with the scipted approach.

In this scenario, a business-critical enterprise sales app slows down. The screenplay specifies the cause, the affected services, the data a user sees at every step, and the exact remediation steps a user would run to fix the issue. Here is how the script beats play out across the two surfaces.

The user starts and ends in the dashboard, uses the chat for investigation, and fixes the problem in the product while the AI guides them.

Impact

Scenario flow for a configuration drift remediation: the script content on the left is broken into beats, and each beat maps to screens in a UX flow across the workload dashboard and Copilot chat.
I partnered with Copilot to develop this situation-based grouping concept. Plausible technical details were essential to envisioning and rendering this.

The experiences shipped and were adopted. The experience is in private preview with four industry-leading adopters: two Fortune 10 healthcare companies, a top-3 gaming and entertainment company, and a multibillion-dollar financial services firm.

They became showcase material. The experiences were featured at major industry conferences by my company and by partner companies.