Einstein Assistant
Designing an AI conversational chatbot for small-to-medium business teams on the Salesforce website. One that assists with wayfinding and personalized product recommendations while balancing trust, transparency, and privacy.
Einstein wasn't a generative AI chatbot. It just looked like one.
When we started this project, Salesforce's chatbot was still called Einstein Assistant. It could produce simple responses and push users to contact sales, and that was about it. No memory, no personality, no product discovery. Users found the experience frustrating and unhelpful.
Endless scrolling with no way to revisit earlier parts of the conversation. Users lost their place and gave up.
After most responses, the chatbot pushed users to contact sales before they had enough information.
Generic, repetitive phrases made the experience feel robotic. No distinctive personality or human-like tone.
Product suggestions were vague and generic. No links, no comparisons, no way to make progress on a purchase decision.
Personality & Tone
We tested three personality archetypes (playful, formal, empathetic) across 16 user interviews using ChatGPT. Participants prioritized content relevance over personality in product browsing, but wanted empathy in troubleshooting and opinions during discovery. The right tone isn't one style. It's knowing when to switch.
Users favored the playful personality for its salesperson-like approach and personal connection. The formal personality provided the most structured responses. The empathetic personality was preferred during troubleshooting.
Context Awareness
We ran 3-5 contextual inquiries across ChatGPT, Amazon Rufus, and Salesforce Einstein to understand how users navigate chatbot conversations. Five pain points kept surfacing: threads and journeys, prompts, reliability and transparency, personal information handling, and length of entries.
Users struggled to find previously discussed products, couldn't start new conversation threads intuitively, and felt uncertain about whether the chatbot understood short entries. They needed navigation, not just answers.
Two tracks. One character. Three principles.
The personality research told us how the chatbot should talk. The context research told us what it needed to do. We merged them into a character (Fin, a dolphin) and three design principles that governed every feature decision.
Fin is curious, empathetic, informative, and engaging. They use human-like language, provide structured responses, take initiative with suggestions, and adapt their tone to the situation, warm and approachable but always professional.
Assist, don't redirect.
Connecting to sales is a feature, not a fallback. It appears only after repeated failed attempts.
Adapt the tone to the moment.
Product browsing gets opinions. Troubleshooting gets empathy and brevity.
Make the information findable.
Every response should be locatable, comparable, and saveable without scrolling the entire thread.
Four features. Two jobs.
Everything the chatbot does falls into one of two categories: helping users find and evaluate the right product, or having a smarter conversation that reads context and adapts. The features aren't independent. They form a single product discovery arc.
Navigate. Compare. Decide.
Two features that solve the same problem: users couldn't find what they'd already seen, and couldn't compare what they'd found.
Chat Timeline
A timeline alongside the chatbox generates clickable headings for each section of the conversation: user needs, product suggestions, comparisons. Clicking a heading jumps directly to that part. We tested two designs: a dark tab (hard to find without onboarding) and a skeleton timeline with a clock icon (preferred, but users clicked instead of hovered, so we switched to click-to-expand).
Ask. Adapt. Re-engage.
Two features that solve the same problem: the chatbot couldn't gather context or respond to hesitation.
Clarifying Questions
Rather than pushing users to agents, the chatbot asks natural clarifying questions (company size, industry, desired features) to gather context conversationally. For brief or vague entries, it confirms understanding before proceeding. This replaced the old approach of routing users to sales before they knew what they needed.
Four participants. Three features. The findings that changed the design.
We conducted usability testing with four business professionals on mid-fidelity wireframes across the timeline, behavioral triggers, and product comparison flow.
Hover doesn't work for timelines.
Users' first instinct was to click, not hover. The hover-to-expand was accidentally triggered repeatedly. We switched to click-to-expand.
Comparison tables need breathing room.
Information was too tightly packed. Users wanted more filter options, customizable categories, and clearer headings.
Button labels create confusion.
Labels like 'Message' confused users. Did the chatbot receive a message, or is this a messaging feature? We rewrote every label.
Reassurance needs a threshold.
Some users found unsolicited reassurance unnecessary. The final design triggers it only after sustained backspacing, not brief edits.
What four months with Salesforce taught a team of nine.
Stakeholder communication is a design skill.
Presenting to Salesforce sponsors every few weeks forced us to articulate design rationale clearly. The ability to defend a decision without getting defensive.
Autonomy is the actual help.
Our instinct was to make the chatbot more proactive. The research kept telling us the opposite. Users wanted to explore independently.
The timeline concept has more to give.
It's still nascent. Deeper research, media integration, returning customer flows, and input from actual sales agents would all refine the experience further.