Case study · Sprinklr

Reimagining UX with Agentic AI

An AI co-pilot that lets enterprise teams hand off what they want done, and get there faster.

A smiling support agent with a headset in a gradient ring, with AI assistant and chat icons

Problem and role: Sprinklr ran on static, point-and-click screens, so every goal became many manual steps. I led the interaction patterns and the AI co-pilot design language.
Team and scale: Lead Product Designer, one of 6+ designers on this work, within Sprinklr's 25+ person design team. Sept 2024 to Mar 2025. Live.
The idea: Keep the AI inside the workflow, and turn its advice into one-click actions.

The Sprinklr AI co-pilot panel open over a listening dashboard, with the greeting card in front — fig. 1 The co-pilot, open in the flow of work.

Outcome

40%

time saved on top-of-funnel tasks

95%

task success, against 75% in manual flows

+25%

NPS after launch

A separate AI page breaks focus. Help should arrive where the work already is.
The principle behind the co-pilot

Decision 01 · Placement

Inside the workflow, not a separate destination

The co-pilot sits as a quiet pulse in the corner and opens into a side panel in one click, without taking people away from the task in front of them.

Co-pilot side panel open over a Sprinklr listening dashboard, answering a question about a product launch — fig. 2 One click from anywhere.

Decision 02 · Action

One-click actions, not just answers

Users lose momentum when they have to turn advice into steps themselves. The co-pilot suggests the next step and lets them take it in one click.

Today’s Updates view: a 32% TikTok reach spike with suggested next questions and tasks beside the co-pilot — fig. 3 An insight, the next question and the task, all in one view.

Decision 03 · Character

An orb, not a face

Robot faces and sci-fi gimmicks work against trust. We chose the orb, a simple form that responds through motion and colour, so the AI feels human without pretending to be one.

Orb motion states on a soft Sprinklr gradient: speaking, listening, on hold, the chat and thinking animations, and nav header states — fig. 4 Speaking, listening, on hold and thinking.

Early explorations of the AI interaction, including a rejected direction labelled too crystal ball magic feel — fig. 5 Early explorations. We dropped the “too crystal ball” directions.

Decision 04 · Brand

Extend the brand, don't compete with it

The co-pilot needed warmth without pulling attention from the product. We extended Sprinklr's visual DNA with a friendly character and motion that guides, in a tone that stays clear and humble.

Co-pilot greeting panel: Good Morning Emily, how can I help you save time today, with suggested prompts — fig. 6 A warm greeting in Sprinklr’s colours, never a separate brand.

Reflection

What I learned

Context makes AI useful

Without the user's context, even a smart system becomes noise.

Protect momentum

Every step between insight and action costs energy. Remove the steps.

Trust is quiet

Visible but never demanding is what makes people rely on it.

Next case study

Actionable Insights