Demo Assist Agent

Demo Assist Agent brings AI directly into the SE workflow, streamlining org setup, personalisation, and storytelling all without leaving the page.

Project Overview

Demo Assist Agent is a conversational AI assistant embedded in the Salesforce Sidebar, purpose-built to help Solution Engineers create high-impact demo environments with minimal effort. From custom objects to branded orgs, it automates the tedious so SEs can focus on the narrative.

As the UX lead, I designed the full end-to-end experience: from conversation flow and visual UI, to feedback collection and launch communications.

Designed org expiration into agent flow to reduce $25M/year in Data Cloud demo org costs
Achieved 20% reduction in infrastructure costs post-launch
Reinforced good demo hygiene through smart nudges and conversational design

Client

Salesforce

Role

Product Designer

Project Type

AI agent Design

Project Timeline

2025, 3 Weeks

The Challenge

Demos are one of the most powerful tools in the SE workflow, but building them is time-consuming, repetitive, and inconsistent.

Despite existing tools, SEs still struggled to:

Spin up industry-specific orgs quickly
Personalize branding and data
Manage expired or cluttered orgs

To complicate matters, early AI prototypes revealed that the agent behaved like multiple assistants, creating confusion and trust issues. Our goal was to design a unified, human-centered experience that felt like a single, smart teammate, not a patchwork of scripts.

User Research:
Surveys & In-Field Feedback

We surveyed SEs globally and conducted deep-dive interviews with power users across EMEA, APAC, and AMER.
Insights revealed:

Over 35% found scattered information the most difficult part of locating resources
Many didn't trust AI tools without transparency
SEs valued control as much as automation

We also gathered testimonials from early testers to validate impact and drive adoption.

Understanding the User

Solution Engineers working at Salesforce
act as a bridge between the technical and sales teams, helping clients understand and implement technical solutions that align with their business goals using product demos and presentations.

“I have dreamed about SEs having the ability to localize and translate demo org data”

Solution Engineer, France

Clélia

Goals:
Speed, control, extensibility

Pain Points:

Redundant clicks, too much setup. Translating metadata.

Solution Engineer, Graduate

Mercy

Goals:
Step-by-step guidance

Pain Points:

Overwhelm, unclear instructions

Principal SE, EMEA UKI

Jim

Goals:
Scale, governance, visibility

Pain Points:

Lack of standards, cleanup burden

Workflow Mapping & Sidebar UX

Working alongside Shriya, who was designing the Sidebar Experience that the agent would live in, I helped define the agent's journey across the Sidebar lifecycle.
We designed entry points, fallback states, and confirmation screens to create continuity from initial trigger (Q Branch app or Sidekick) through to asset install and org branding.

Key UX moments included:

Smart suggestions based on org type
Quick installs with contextual links
Light feedback capture after each task

Conversational UI & AI Design

We designed the agent to feel reliable, transparent, and helpful — never overconfident.
Here’s how we built trust and utility into every step:

Citations & Source Transparency
Every recommendation or install included a clear source, description, and preview.

Transparent Disclosures
The agent clearly stated what it could and couldn’t do — no hidden automation. We done this through transactional emails, as well as being embadded into the agent.

Feedback Components
Mini prompts after tasks helped gather user input while signaling responsiveness.

Model Containment
Instead of relying on open-ended generation, we scoped intents to pre-defined Salesforce tasks. This reduced hallucinations and made behavior more predictable.

Branding

The branding for Demo Assist Agent was designed to feel like part of Salesforce’s Agentforce ecosystem while visually reflecting the tool’s unique role as a behind-the-scenes enabler for Solution Engineers.

Name Change: From Agent 007 to Demo Assist Agent

The project began as Agent 007 — a nod to the stealth, precision, and utility of the experience. As we matured the product, we shifted to Demo Assist Agent to better communicate its purpose while maintaining the agent-like persona. The updated name aligned with other Salesforce assistants (like Sales Assist and Service Assist) and clarified its connection to demo workflows.

Logo & Visual Identity

The logo combined the official Agentforce icon with a barrel design motif inspired by classic James Bond intros — tying back to the tool’s original "Agent 007" code name. The circular barrel graphic gave the logo a sense of motion and precision, reinforcing the idea of a smart, targeted assistant embedded within the Sidebar.

Design considerations included:

Consistency with the Agentforce design system (colors, iconography, shape language)
Recognizability at small sizes within the Sidebar and Slack
Narrative alignment with the product’s identity as an intelligent, efficient agent working behind the scenes

The result was a visual identity that felt playful but professional, distinct but fully integrated within the Salesforce AI design language.

User Testing

We conducted both in-person and remote user testing across multiple SE regions to validate the conversational flow, task completion, and overall experience.

To streamline continuous feedback, I created a custom Slack-based feedback workflow:

After each interaction, users received a short feedback prompt within Slack
Responses were automatically logged into a connected Google Sheet
We then used Claude (AI) to synthesize qualitative insights and detect sentiment patterns across responses

This workflow allowed us to gather real-time feedback during the pilot phase without burdening SEs or requiring formal survey cycles.

Key outcomes from testing:

SEs appreciated transparency around what the agent could and couldn’t do
Users requested clearer language when confirming installations or branding steps
The single-agent flow (vs. multiple personalities) significantly improved trust and task clarity
Bug identification

This process helped us quickly iterate on tone, flow, and functionality before broader rollout and laid the foundation for a scalable feedback mechanism as usage grows.