Interaction Design User Research

Enterprise Chat

Redesigning how SAP's Digital Assistant and Human-to-Human chat coexist on Fiori Launchpad — resolving user confusion between two overlapping communication products through systematic exploration and validation.

Timeline

3 months

Focus

Interaction Design

My Role

Lead Designer — owned end-to-end design from secondary research to design validation, collaborating with one other designer.

The Challenge

Two chat products had coexisted on SAP Fiori Launchpad for years: a Digital Assistant for submitting requests, and Human-to-Human Chat for peer communication. They had been coupled and decoupled multiple times, and were currently placed under the same header — causing persistent user confusion.

🎯

Problem Statement

How should Digital Assistant and Human-to-Human Chat be positioned to coexist on Fiori Launchpad so that it's easy to understand?

The scope was specifically about placement and layout — not the functionality of either product.

The existing design showing DA and H2H under the same header

The existing design — DA and H2H under the same header

Design Highlights

Separate Entry Points

Users access the Digital Assistant from a floating button on the bottom right, and Human-to-Human Chat from the navigation bar. Separating the entry points immediately resolved the confusion between the two products.

Design decision: separate entry points

Distinct Panel Behaviors

The Digital Assistant opens as a floating panel on the bottom right, while Human-to-Human Chat docks to the right side of the page. The different panel styles reinforce that these are distinct tools with different purposes.

Opening chat windows animation

Stacked Windows Option

For users who want to conserve screen space, we provide an option to stack the windows as a transient layout. This respects power users' workspace preferences without compromising the default clarity.

Stacked windows animation

Research & Process

We applied the RITE method (Rapid Iterative Testing and Evaluation) over 2.5 months — generating ideas rapidly, testing with users biweekly, and evolving the design based on real feedback across 4 structured rounds.

🔍

20+ Explorations

Analyzed chat patterns across LinkedIn, Facebook, IBM, Salesforce, Adobe, and Slack — then generated over 20 design explorations covering entry points, panel layouts, and interaction models.

🧪

4 Testing Rounds

Structured biweekly sessions with 4+ participants each round. Each cycle generated both improvements to existing concepts and entirely new directions.

👥

21 Participants

Across all rounds and the final validation survey, 21 users shaped the design — ensuring the solution was grounded in real feedback, not assumptions.

Design Validation

To convince the Fiori design council, we ran a survey with 16 participants comparing three options: the existing Fiori 3 pattern, a Fiori-guideline-compliant exploration, and our proposed design. Our proposal ranked highest in preference, earning support for implementation.

Option 1
DA H2H

Icons in center header — users confused DA with H2H

"I can't figure out the digital assistant icon. Is it the circle in the header?"
Option 2
DA H2H

Icons close together — placement still ambiguous

"It is confusing to see the difference due to the close placements."
Preferred
Option 3
H2H
DA

Separate entry points — instantly clear and discoverable

"Easy to find everything!"

Impact

16
Survey validation participants
#1 preferred
Among all tested options

Takeaways

🔬

Explore Broadly, Converge Deliberately

The RITE method forced us to resist premature convergence. Early rounds produced surprising insights that fundamentally changed our direction.

📊

Data Persuades Stakeholders

Quantitative survey data combined with qualitative user quotes gave us the evidence needed to win design council approval for a significant platform change.

🧩

Separation Creates Clarity

When two products share an interface, users conflate them. Distinct entry points and panel behaviors made each product's purpose immediately obvious.