Grabbing pixels...

Using AI to sway the Pendulum

Pendulum case study cover

My team and I reimagined how AI could be used to simplify the risk management workflow, enhancing Pendulum.AI’s capabilities by increasing usefulness by 40% and usability by 80%.

Team: Me (Design, PM), Alex (Research), Sai (Design, Accesibility), Yani (Research)

Duration: 7 month Sprint

1. INTRODUCTION

What is Risk Management?

Pendulum's proprietary AI technology is designed to provide tech companies with real-time insights into who is discussing their brand, across what platforms, and at which channels – enabling proactive and effective reputation management.

Their main business goal was for us to build a feature that would help them stand out from their competitors. We also had to use their highly-trained AI models in a unique way.

2. USER RESEARCH

We engaged with executives from Pendulum’s customer portfolio (Zoom, T Mobile, Google, Amazon, etc.) and industry leaders to explore design opportunities in a mixed methods initiative.

COMPETITOR ANALYSIS

10+ AI driven data platforms like Servicenow, Onetrust, Done Safe, etc to understand industry trends.

eXPERT INTERVIEW

4 Dis/Misinformation experts to get nuanced understanding of content moderation, its common practices, pain points, and aspirational expectations.

QUALITATIVE RESEARCH

3 Journey Maps and 4 Contextual Walkthroughs with clients to get tool specific expectations of real users and understand the basic usability of the platform. 5 Internal team interviews to understand business context.

80% of data points that users engage with in 10 mins have sensitive content. Features to support mental fatigue of an outdated and clunky dashboard and distress of seeing sesitive content was of huge interest.

Emotional journey map

3. DESIGN SPECS

Design Principles and Goals: Since these were experince derived from users' emotional journey we the tool, grounding our vision through these specs helped drive the process.

We brainstormed several ideas and conducted SWOT and impact matrix analyses to narrow down our focus. We also asked business leaders and engineers to engage in a SWOT analysis workshop.

4. OKR 1

I wireframe and conducted concept evaluation for the different scenarios where users could utilize an AI assistant in the main dashboard experience. 40% of the users were hesistant because the need for automation should not supersead their need for autonomy. In my next iterations, I focused on providing a flexible experience and safe.

USERS

40% of the users were hesistant because the need for automation should not supersead their need for autonomy. In my next iterations, I focused on providing a flexible experience and safe.

BUSINESS AND ENGINEERING

Our VP quoted  "if it is not broken, why are we fixing it" but engineers saw real value in cleaning up the filtering and tagging system and using the exisitng AI in clever ways. I worked with the VP to build a cautious testing timeline for this feature.

GenAI Copilot

The co-pilot made data analysis and reporting easier. Users no longer have to build a complex filter group to find information quickly. Additionally, this reduces mental fatigue and encountering toxic content.

5. OKR 2

My other designer wireframed over 4 ideas to for the content blurring feature. The biggest challenge was to not hinder with their productivity. My main role was doing design review.

+ Does it align with research data

+ Is the UX intuitive and simple

+ Are we providing business value?

+ Does it align with the design system

+ Are all the edge cases considered?

Contextual Search & Blurring

Sometimes, toxic content cannot be avoided. Contextual blurring helps users get a quick summary of sensitive content without actually having to see it. Since it is a filterable toggle, it is designed to not hinder with everyday work.

Content blurring final screen

5. OKR 3

I wireframe and conducted concept evaluation for 3 ideas where users were reminded to take a break and decompress from seeing sensitive content. The biggest challenged was not being tacky and overbearing for the user.

A wholistic mindfullness space to wrap up all the features. You can look at your footprint in Pendulum and access resources to make your work life easier.

Give a shout-out to a coulegue you think needs it. Can be provided publicly or personally to help keep up the team morale for such a taxing role.

Take quick and regular break reminders. Customize when you want the system to remind you to refresh. Select from a list of on-device and off-device activities to do.

Mindfullness Space

Introducing the mindfullness dashbaord - get break reminders, resources, collaborative tools, and impact tracking. Also browse how you can reduce mental fatigue and distress while Pendulum.

Mindfullness feature final screen

06. FEEDBACK AND THOUGHTS

The team did a great job in understanding the requirements and delivering something unique yet achievable. A lot of people loved their work (especially Reddit's Design Leadership)

-Nikolay Tradev, VP, Pendulum

My Thoughts

Designing AI-driven tools requires balancing automation with user control, ensuring trust and adoption, especially in specialized B2B environments.

Rapid iteration in a startup environment demands a deep understanding of user needs, and the ability to quickly pivot based on user feedback

My team

Also check out..