Salesforce
Chatbot
A Salesforce sponsored project
What was asked of us
Design a GenAI-powered chatbot for Salesforce that effectively communicates and efficiently addresses user needs.
With a focus on Smooth interactions, quick results, and reinforcing trust and security at every step.
Focus areas
1. Contextual Awareness
2. Tone & Personality
3. Product Discovery
Research
Contextual Awareness
We conducted probes with 5 participants with a business background to reflect Saleforce's ideal audience.
Take a peek at the detailed research insights here
Our findings
We observed that the users had difficulty with the following sections:
Length of Entries
-
Generally used single worded entries.
-
Entries are sometimes very short and not very detailed.
Prompts
-
Struggle to structure potential queries effectively.
-
Insufficient follow-up/clarifying questions.
-
Prompts push users toward agents, limiting independent exploration.
Threads /Journeys
-
Difficulty starting and managing new threads.
-
Scrolling back and forth to refine or correct entries.
-
Lack of filters and sorting for concise, relevant responses.
Product discovery
-
Product details presented in plain text lack organized points.
-
Losing track of conversation in lengthy chats
-
No efficient way to compare products.
Fear of closing chat
-
Not enough signifiers to inform what happens when closed
-
feel the need to browse the website while chatting.
-
Fear losing product links shared by the chatbot
Sketches
Research
Tone & Personality
We designed 3 distinct personality types and tested them on various participants to test which one the users prefer with the intention of browsing/buying salesforce products.
Take a peek at the detailed research insights here
Our findings
Through our various probes we found that the users preferred :
Tone Preferences
-
Formal Tone : Preferred for workplace interactions because it is professional and structured.
-
Empathy in Troubleshooting : Users emphasized the importance of empathy during troubleshooting.
-
Adaptive Tone : Some participants liked the idea of the chatbot detecting frustration and changing its tone to better suit the user’s mood.
User priorities
-
Participants valued relevant and accurate content more than the chatbot’s personality or tone.
Perceptions of the Chatbot
-
Some participants found it hard to imagine who they were talking to since it’s an AI.
-
Others imagined they were talking to an older person.
Personality Preferences
-
Fun Personality : Favored for its personable, salesperson-like approach and personalized recommendations.
-
Formal Personality : Preferred for delivering clear, structured responses, especially in workplace scenarios.
-
Empathetic Personality : Appreciated for its sensitivity but needs enhanced empathy during troubleshooting.
Ideation
Final Features
We decided on a few main ideas that will be triggered based on user behavior which addresses most of the issues we faced during our research phase. A few of my favorites are listed below:
1. Chatbot Mascot
2. Timeline
3. Comparing Products
4. Clarifying questions & Suggestive Prompts
Final Product
Chatbot Mascot
We designed our own new representative for the Salesforce AI chatbot using their design guidelines. We chose Fin the Dolphin because it closely matched with the characteristics of our bot, which were:
%203.png)
Fin the Dolphin
Timeline
Users can keep track of chatbot conversation based on topics of entries. And jump to a specific section to find relevant responses they need.
Ideation
Product discovery
This focus area mainly consisted of ideation because all our research for this was gained from probes that we conducted during our contextual awareness research phase.
Comparing Products
We added product cards that are easier on the eyes and convey information better. As well as compare products feature where you can save the summary if you wish to do so.
Clarifying questions & Suggestive Prompts
Clarifying questions that help the bot understand the users much better without overwhelming the user to give out more information like their email.
Suggestive prompts help the user frame their question better incase they get stuck in a spot trying to generate a prompt.
