Panaseer Key Drivers
The Key Drivers feature is an AI-powered personal investigation assistant. It empowers security analysts to quickly find the reasons behind a metric change, so they understand not only what happened but why the number dropped, and can close security gaps faster. Since it was Panaseer's first AI-powered feature, as a lead designer on this initiative, I first facilitated the Future of AI workshop with our CTO, CPO, lead data scientists, and lead engineers. To gather feedback on my prototypes, I then ran user research with 20+ security stakeholders from organisations including AXA, BlackRock, WPP, and MUFG. What surprised me most in usability testing was that most users preferred natural-language explanations to graphs or tables. They could understand the change and copy it into their reports much faster.
CONTEXT
Security analysts bounce between hundreds of metrics and dashboards just to work out why their numbers changed. Investigating a single shift in risk posture can take days, which is overwhelming, and teams often fall back on assumptions instead of evidence. This kind of analysis is well suited to AI and valuable to Panaseer users, so we chose it as our first AI opportunity
Key Drivers shipped to all Panaseer customers as the first product in the Panaseer AI suite. The insights from it led directly to the next Gen-AI features- MetricIQ and, finally, the Cyber Advisor as well.
INSPIRATION AND MOTIVATION
My approach was “AI-ASAP”. When I started, Panaseer had no AI features, while the market around us was filling up with AI-driven products. I didn’t want us to fall behind, so I strongly advised our leaders to bring AI into the product as soon as possible. Rather than wait for a perfect long-term AI strategy, I pushed to start with one clear, valuable problem: slow, manual investigation. The question that guided us was: How might we help security teams spot and close their control gaps faster, at a time when attackers are also using AI to move quicker? Answering it started with helping analysts understand why their metrics change. That first step became Key Drivers, and it opened the door for everything that followed.
OPPORTUNITY
From opting out of AI to asking for a natural language interface
In late 2024, when we first started exploring AI, our information security users told us they weren’t using AI, nor were they ready to start. Many work in the cyber teams of banks and insurance companies, where regulation makes them cautious, while we on the product team were excited about innovation and adopting new AI tools.
To understand what would earn their trust, I followed a human-centered design approach and built one feature at a time with our users. We started with a simple question: how might we help InfoSec teams understand their own data faster, and in particular, why a number of a metric changed? (root-cause analysis). I invited those users into the R&D process so I could learn about their specific pains and needs, and so they could imagine and test how AI could empower their roles. Eventually, I could see a correlation between the collaboration we had and their AI acceptance, as they shared their own ideas with us and even brainstormed AI agents on every page of the complex, data-heavy product - which then became Cyber Advisor.
AI DESIGN SYSTEM STEPPING STONES
In February 2025, I began exploring how best to communicate an AI=powered feature in the product vs. the other features we had. It was important to me not only to innovate and highlight our new AI feature, but also to ensure it aligns with our in-house design system for consistency and trust.
During the research, we noticed that although the experience is familiar and easy to use, a blank prompt box can be intimidating, and users needed a little guidance to get started. To solve this, we added suggested starter questions to the UI, such as “What’s my EDR coverage per BU?”, which lowered the barrier to the first prompt and helped drive adoption even higher.
How might we build user trust in an AI system?
Consistent UI and mental models: “New” can introduce uncertainty. Because this was the first time we designed and developed for and with AI systems, I wanted to keep our UI components as consistent as possible. We also borrowed a familiar interaction pattern from gen-AI products already on the market, so users could rely on habits they’ve formed elsewhere. This meant they didn’t have to think about how the interface works and could focus on the AI’s response. The experience feels intuitive because it takes place in a UI they already know and trust, and the impact was a higher rate of adoption.
PRODUCT DEVELOPMENT
The key learnings that led to the Cyber Advisor feature
MetricIQ insights:
Before we developed the Cyber Advisor, we had 3 different gen-AI features on different pages of the product. I set up many feedback sessions after the release to make sure we built not just another AI tool, but the most valuable and accurate features for our customers. MetricIQ is an AI-powered natural language summary of the most important risks so the CISO can focus on taking rapid action to fix them. My key learning was that our users were missing one big thing: continuous prompt interaction. They wanted to ask follow-up questions and go deeper into their data. That insight became the foundation for Cyber Advisor.
Cyber Advisor everywhere:
During research and prototype testing, one of the most common ideas I heard from our users was that they wanted the AI agent to be visible on every page of the UI. Based on that feedback, right after releasing Cyber Advisor, we added a mini window that appears on every page and understands the context of the page it's on, so users can ask questions wherever they are in the product and get super-tailored information in seconds. As an outcome, the adoption of the Cyber Advisor increased across all users.
Typography in design
Over the past two years, leading design work on GenAI-powered features, I’ve seen a clear shift: interfaces are becoming increasingly text- and typography-heavy, with fewer visual anchors. And with that shift comes a familiar pressure- “let’s reduce the white space to fill the layout.”
However, just like the Japanese design philosophy, I strongly advise my team to let that space exist. It allows the user’s mind to pause and process.
Whitespace creates structure in content-dense experiences. It enables focus, supports comprehension, and helps users to engage with the information presented in the UI.
As AI accelerates the mass production of content (not always high quality, imo), the role of whitespace becomes even more critical. It is one of the few levers we have to restore clarity and reduce cognitive overload at scale.
Sometimes, the best way to enrich a UI and achieve high-quality design isn’t to add more- it’s to give it room to breathe.
