Crystal is a software company that leverages its proprietary AI to predict personalities based on online presence. Their platform provides users with tailored communication strategies to enhance rapport, persuasion, and negotiation.
Project type: end-to-end
Duration: 8 months
Project role: UX research and design
The short and sweet
Recognizing that sales professionals and researchers heavily relied on Crystal for personality insights but faced challenges in comprehensive meeting preparation, I led the design of a tool to streamline pre-meeting research. I conducted user interviews, identified key pain points, and formulated a problem statement highlighting inefficiencies in manual research and data gaps. Through workshops and pilot testing, and in collaboration with our development team, director or research, and head of product, I designed an AI-powered solution offering automated meeting prep, personality insights, real-time data updates, and competitive intelligence, effectively enhancing sales professionals' efficiency and confidence.
User research
Business case
Crystal’s most active users were sales professionals and researchers supporting sales teams, relying on the platform for personality insights on their prospects.
While Crystal’s data improved communication, there was a market need for a more complete meeting prep solution.
The company sought to expand its offering by developing a well-rounded tool that would streamline pre-meeting research and equip users with actionable insights beyond personality analysis.
I interviewed 15 sales reps to learn:
1. About their existing meeting prep workflow:
What types of sales meetings do you typically prepare for? (e.g., discovery calls, demos, negotiations)
Walk me through how you prepare for a sales meeting. What steps do you take?
What tools or resources do you rely on to gather information before a meeting?
2. Identify key frustrations and inefficiencies.
Have you ever felt underprepared going into a meeting? What happened?
How do you determine what’s most important to focus on before a call?
The sales representatives I interviewed:
Have 5 to 8 meetings a day
Spend 15-30 minutes prepping before a meeting
Before meetings they:
Scan LinkedIn, internal CRM, and news about the company
Review notes from previous meetings and competitors
Empathizing with users
The Impact of Poor Customer Insights in Sales Meetings
Generic sales pitches → Low engagement
Missed opportunities to address key pain points
Failure to build trust & rapport
Difficulty handling objections in real-time
Longer sales cycles & lower conversion rates
What we found
Screenshot from a synthesis exercise in Miro
Alex is a busy, results-driven sales professional who struggles with time-consuming meeting preparation and a lack of deep customer insights. He needs an AI-powered tool that quickly delivers personalized, data-driven briefings to help him confidently engage prospects and close deals more efficiently.
The target persona
“It’s frustrating because I don’t always know the best approach until I’m already in the meeting.”
Problem statement
Sales professionals waste time on manual research, lack deep customer insights, and struggle with personalizing engagement.
Incomplete data and missed competitive signals lead to misaligned messaging, lack of confidence, and lost deals.
Prioritized problems based on frequency & impact
Top Priorities for an AI-Powered Solution
Automate meeting prep → AI-generated pre-meeting reports with CRM, LinkedIn, and market insights.
Provide personality insights → AI-driven recommendations on how to approach decision-makers.
Improve CRM data quality → Automatic updates with real-time company and prospect information.
Competitive intelligence in real-time → AI alerts on competitor mentions and industry shifts.
Smart follow-up automation → AI-generated summaries, action items, and email drafts.
Ranked list of our user’s’s challenges, ordered by how often they were mentioned and the impact on their workflows and sales success
User stories
As the team began planning how to build an AI-generated pre-meeting report, I led a workshop for internal stakeholders to define:
What info might be helpful for meeting preparedness?
What data do we have access to, and how can it be synthesized into actionable insights?
Workshopping
Workshop key takeaways
Automating data collection and prioritizing key insights is key to a useful product.
Relevant data could be determined via machine learning to ustomize insights based on meeting context (e.g., prospecting, negotiation, QBRs, etc.)
The tool should help them navigate conversations strategically, applying Crystal’s personality based content to the meeting participants, industry, and type of meeting.
After determining what we could quickly and feasibly produce, the team ran a test pilot with users who agreed to give feedback. This ran for about 6 weeks and allowed us to work through the best ways to gather, synthesize, and display information.
Users were provided PDF reports 2 days before a meeting with preparation content.
What users said:
Waiting for a PDF to be compiled and sent worked for really important meetings, but wasn’t realistic for everyday use.
The reports were way too long (some were over 20 pages) and sales people didn’t know where to focus or what info would help them most.
Users found actionable content most helpful. (“Say this” / “Don’t _____”)
While having info on each meeting attendee was helpful, the real value came from understanding how the group would function together.
Test pilot
“This report is awesome—tons of valuable info here—but honestly, I don’t have time to read all of this before every meeting. I need something that cuts through the noise and tells me exactly what matters most, fast.”
Using AI for ideation
With a clear understanding of the essential UI features, I experimented with AI-generated wireframes to explore the technology’s capabilities. While still in its early stages, AI helped accelerate the initial ideation process and provided a solid starting point for the UI. Using progressively refined prompts like the one below, I generated 10 variations to see how AI would interpret and structure the design.
Initial ideation
Testing iterations
Next I conducted user testing through two methods: online testing platforms and virtual calls with real users using a Figma prototype. In both scenarios, I asked users to click through the prototype to perform various tasks, such as connecting their calendar, locating their meetings, viewing the premeeting report, and locating information within it.
User Testing
Users were generally able to connect their calendars. I made sure to follow industry standards for integrating with 3rd party integrations.
The dashboard was overwhelming for some. Those with fewer meetings had an easier time navigating, likely because their screen has less info displayed.
Finding the reports was not easy for users - even people on our own CS team! This was clearly a piece that needed more attention.
The sheer volume of information in the report was still overwhelming to users. They needed to sort through the clutter to find the gold.
Our developers used machine learning to automatically tag each meeting with a “type”. This allowed us to tailor the report content to be more relevant for each meeting.
Additions based on testing
In this example the meeting has been tagged as “Sales meeting”. This serves as the fall back when we weren’t able to determine a more specific meeting topic, and the report will be the most content dense. Users are always able to edit the tag manually to see a drilled down version, or use the quick tips section to generate content.
Here the meeting has been tagged “Pricing discussion”. The layout still has some global content, such as meeting attendees, but the key advice will be tailored towards budget discussion and tips for negotiating with the personality types in the meeting.
Unlike most of Crystal’s existing features, the mobile experience for PreMeeting had significant importance. This is because many users prepare for their meetings before they reach their desk. Some said they checked their calendar before going to sleep so they knew what to expect the following day, while others told me that any prep work they did was over their morning coffee. I did two key things to account for this:
Daily emails that summarize the user’s meetings for the day. These can be customized to be sent at what ever time the user prefers.
Direct links to the report, with an extremely mobile-friendly view.
The importance of mobile
Final designs
Initial connection
Connecting your calendar is seamless and quick by integrating directly with Gmail and Outlook’s SSO.
Once users have connected their calendar, their meetings and links to attendees profiles appear in the dashboard.
They can customize their preferences with options like “Don’t create reports for meetings with only internal attendees”
Users select the time they want their report summary to be emailed to them -that way it is up to date and in their inbox before their meeting preparation.
Meetings dashboard
Dashboard provides at-a-glance view of your daily meetings and quick links to attendees profiles
Upcoming prospects section lists attendees from the day’s meetings. Clicking a person opens their in-depth personality profile
Prepare now buttons open the PreMeeting report with a customized set of information based on the identified meeting type
PreMeeting report
Meeting attendees were moved to their own column that persists for actionable advice. This is more useful for cross-reference or use during a meeting
Quick tips allows users to generate specific content without parsing the entire report
Information most relevant to the meeting type is found in the right column. Users are able to change the meeting type as-needed to change the content of the report
Use the full prototype below:
For a short time, the team considered a “Wildberry Poptart” themed design. It was around the holidays… maybe we weren’t thinking straight. While we didn’t end up going in this direction, it was a fun exercise in dark mode theming!
Bonus
Once again, I leveraged AI for initial testing, focusing on composition and clarity to ensure the dashboard and report were intuitive. This helped refine the design, guiding users toward the desired actions.