IBM Client Engineering — Fortune 500

Making Preethi

Designing a live conversational AI avatar — from blank brief to standing ovation — in three weeks.

Role

Sole Digital Artist, Co-UX & Conversation Designer

Audience

150+ Executives across 2 conferences

Team

IBM Client Engineering

Timeline

3-week sprint

The Challenge

How do you make a room full of skeptical executives believe in the future of AI?

A global energy company came to IBM Client Engineering with an unusual request: create a "wow factor" experience for their upcoming executive leadership conference. Not a slide deck. Not a demo reel. Something live, something visceral — something that would make 150+ senior leaders feel the potential of conversational AI rather than just hear about it.

The ask was simultaneously thrilling and terrifying: build a real-time, human-like AI avatar that could hold a conversation on stage, embody the company's values, and survive the scrutiny of a C-suite audience — in three weeks.

The core design question

How do you design a digital person — not just a chatbot — whose personality, voice, and presence feel trustworthy enough for an executive stage, but human enough to create genuine connection?

My Role

The person responsible for making an AI feel human.

I was the sole Digital Artist and one of two UX Designers on this sprint, leading everything from stakeholder research through final personality design. That meant I owned the decisions that mattered most to the audience experience: who Preethi was, how she sounded, how she moved through a conversation, and what she communicated about the client's vision for AI.

I also served as the connective tissue between the engineering team building the underlying watsonx and Soul Machines integration and the client stakeholders who had strong opinions about brand, culture, and tone — and not always the same opinions.

Process

Three weeks from blank page to live stage.

1. Stakeholder Research in a High-Stakes Environment

Ran rapid discovery interviews with client stakeholders to understand not just what they wanted Preethi to do, but what they needed her to represent. In a conference environment, every design choice — her name, her accent, her sentence structure — carries cultural and political weight. I mapped those constraints before touching a single design decision.

2. User Persona Development → Avatar Persona Design

Synthesized stakeholder interviews into three distinct user personas representing the executive audience — their expectations of AI, their communication styles, and their comfort thresholds with novelty. From those personas, I developed Preethi's personality: her voice cadence, vocabulary level, warmth-to-authority ratio, and how she handled uncertainty gracefully rather than robotically.

3. Conversational Flow Design

Designed the full interaction model for live audience use — mapping the paths a conversation could take on stage, identifying where Preethi needed to redirect gracefully, and scripting fallback behaviors that felt natural rather than broken. The challenge was designing for an audience that would try to trip her up, not just use her as intended.

4. Iterative Voice & Expression Tuning

Using Soul Machines, I customized Preethi's facial expressions, vocal tone, pace, and accent through repeated live testing sessions. This was UX research in real time: I'd run a sample interaction, observe where the experience felt uncanny or cold, and adjust. We did this until the experience felt less like a demo and more like a conversation.

5. Live Conference Deployment — Twice

Preethi debuted at the client's first executive conference and was received well enough to be featured again at a second event. Between the two, I incorporated feedback from the first run to refine her responses and expand her conversational range. Each conference was a live UX test at scale.

Key Insight

"Executives aren’t wowed by technology that doesn’t work, not even once. There can be no slip-ups in live demos or you lose them."

— Cyrus Pilling, reflecting on designing for executive audiences

The biggest design lesson from Preethi wasn't about AI — it was about trust. Executives from well-established, high stakes industries, such as Energy companies, don’t take chances with their investments. The moment Preethi fails to answer a question in an intelligent, interesting, and insightful way, she'd lose the room. The moment she felt like a genuine, helpful presence is the moment this demo gets our team projects with C-suite level clients at global Energy companies.

That meant every design decision — the slight pause before she answered a complex question, the way she acknowledged not knowing something — had to be intentional. Designing for executive audiences taught me that the bar for human-centered AI isn't just "technically impressive." It's "do I trust this?"

Outcomes

From prototype to standing ovation.

🎤 Two Live Executive Events

Preethi performed live at two separate high-profile leadership conferences for 150+ executives, including the CEO — with zero critical failures during either event.

📈 Increased AI Investment Interest

The experience directly increased stakeholder appetite for customer-facing AI applications, shifting the conversation from "should we explore this" to "how do we scale this."

🏗️ Reusable AI Persona Framework

The persona design process I built for Preethi became a reusable framework adopted by other IBM Client Engineering teams for future conversational AI engagements.

⭐ Internal Recognition

Received formal internal recognition from Americas Sales leadership for delivering an experience that balanced technical sophistication with human-centered design — a rare combination on a 3-week timeline.

Methods & Skills

What this project required

This sprint demanded a different kind of UX thinking — one where the "interface" was a personality, and the "user testing" happened in front of a live executive audience.

Conversational UX Design

AI Persona Design

Stakeholder Research

User Persona Development

3D Character Modeling

Live Event Design

IBM watsonx AI

Human-Centered AI

Soul Machines

Executive Communication

Next
Next

UX at Scale: Optimizing Enterprise File Management