For nearly twenty years, Martin Louis has been at the forefront of software engineering and artificial intelligence. Now a Senior Engineering Manager at PayPal, a patent holder, and advisor to AI startups, he has witnessed firsthand how rapidly advancing technology is reshaping the way companies understand their customers. In this interview, Louis reflects on how behavioral analytics and AI are converging to transform personalization, fraud detection, and the digital economy and what that means for the next decade of innovation.
Q1. With nearly two decades of experience in software engineering and AI, what excites you most about the convergence of behavioral analytics and artificial intelligence today?
What excites me the most is that the computational power we have today with GPUs is an enormous advantage to anyone who is analyzing customer behaviors over a period of time. When I started with personalized servicing and search results back in the day, the ROI of storing such behavioral data was hard to justify. However, with everyone expecting customized service, it became clear that customers want big companies to know more about them and provide services that matter to them, tailored to their lifestyle. For instance, until Google Ads started to showcase the power of web analytics. No one ever thought of personalization based on the historical behaviors of customers online.
The cheaper cost of storage, competing power, and AI developments are helping us analyze data in ways that were previously impossible. Long-term and short-term intent detection became a breeze with AI LSTM models. Training and re-training became cheap, and prediction accuracy increased – the more we can train these models, the more accurate they become. We have now reached a point where we can predict future user behaviors fairly accurately and analyze trends in multiple dimensions in almost real-time.
Q2. How do you see real-time behavioral interpretation changing the way organizations approach personalization, fraud detection, and operational intelligence?
The shift is dramatic. We are blessed with computing power and algorithms that can process streams of real-time data; this means businesses can be quick in providing personalized services and offers and have better operational intelligence.
Organizations can leverage a lot of behavioral data generated by their customers on their products and outside on devices like their phones over a longer period of time to provide personalization.
The future in the personalization space is already shifting with MCP servers and agent providers like ChatGPT and Perplexity having a lot of historical context about your engagement with the agent itself . This space is going to open a new door to how those agents are getting hyper-personalized and how that context will be shared with other companies if they want to leverage that interaction history and context. Already, a massive shift is in motion.
Fraud is a slightly different tangent – as both sides are benefiting in good and bad ways, we start to see advancements in ways technology is used by fraudsters at scale. It’s getting harder and harder to find fraud and deeply layered attacks using artificial intelligence. At the same time, cybersecurity teams are also leveraging advanced computing and AI to constantly up the game of fraud detection and risk mitigation.
Behavioral data and digital signatures are helping a lot – but at the same time, we are seeing more synthetic generation of those digital signatures being generated by the latest Gen AI models. This is making the cybersecurity space very interesting and challenging.
Q3. You’ve led engineering and innovation at companies like PayPal. Can you share an example where understanding user behavior at scale made a measurable business impact?
While I can’t share the specifics on how we implement things at PayPal , as I mentioned before , customers today expect brands like PayPal to be aware of their needs and provide what they want when they want .
Global brands are already trusted by customers to hold their personal and financial data – so they expect these companies to use as much data as they can about them and provide services that matter the most to them, and not treat them as a generic user.
Today’s customers expect large companies to recognize and understand them in every interaction. For example, imagine walking into your bank and having the branch manager greet you by name, then mention the new card you signed up for last month, along with a relevant offer. You’d likely appreciate that level of personal attention. Online customers have the very same expectation when they log in to their bank’s website or mobile app.
Q4. Building systems that transform raw data into actionable insights is a significant challenge. What are the key architectural or design principles you believe are critical for success?
Some believe that building a unified data lake is the first step, but in today’s world of modern cloud infrastructure, that isn’t always necessary. It’s perfectly fine for data to be federated or spread across systems, as long as it’s well-structured and easy to understand. The key is to make your data defined and cataloged to a level where even an AI agent can interpret it confidently.
The next step is to establish a central product knowledge base. High-quality data on user behavior is essential, and it can be even more powerful when enriched with contextual facts about the user.
Alongside this, you should build an alert and issue-tracking data set, capturing the health of your systems. For example, if a user’s behavior shows repeated retries, you need to know whether it was due to their actions or simply because a server was down.
Finally, incorporate service and operational touchpoints from the user journey. Once these elements are in place, you can feed the combined information into AI models to generate insights, detect anomalies, identify churn patterns, and deliver personalized services.
Q5. As AI models become more advanced, how can we ensure behavioral analytics respects privacy and builds user trust while still delivering value?
Trust must be engineered into the system; explainability is key to building trust in AI systems, but it all starts with making customers feel empowered about how their data is collected and used.
- Transparency: users should know what is collected and why. Hidden tracking is a fast way to erode trust.
- Consent and Control: empower users to manage their behavioral footprint, to opt in or out in meaningful ways.
Every time a notification or an action is taken on a customer account due to behavioral data being processed by AI, we have the obligation to explain the reason why we came up with this offer.
Example when you look at Netflix – you will see movie recommendations, but will also explain why .. “because you watched x, we recommend y “
Q6. In your view, what role will generative AI and natural language processing play in making behavioral analytics more accessible and actionable for decision-makers?
Generative AI is already transforming large-scale data analytics. Today, LLMs can translate natural language questions into SQL queries that run directly against a datastore. This seemingly simple capability has powerful implications, making analytics far more accessible. Executives can ask plain-language questions and receive immediate insights, without needing technical expertise.
This shift means executives, marketers, and product managers can explore behavioral data freely without knowing SQL or relying on data science teams.
Q7. You’ve also worked on NFTs and e-commerce. How do you see behavioral analytics shaping the future of digital marketplaces and emerging technology ecosystems?
Digital marketplaces face the same core challenge as traditional commerce platforms: building trust. Managing risk and detecting fraud requires understanding the user without compromising their anonymity. Blockchain technologies make this possible by enabling the use of behavioral data rather than personal identity data.
- Ecosystem for Trust — Spot fraudulent listings or toxic community behavior by detecting subtle behavioral patterns.
- Ecosystem for Relevance — Connect buyers with authentic sellers through contextual behavioral models, moving beyond simple keyword matching.
- Ecosystem for Engagement — Frame digital assets, such as NFTs, as part of a user’s identity and journey, not just as static items in a wallet.
Q8. What advice would you give to technology leaders and engineers who want to build platforms that don’t just process data but interpret human behavior to drive better outcomes?
I’d offer four pieces of advice:
- Stay Curious — Behind every click, swipe, or pause is a person. Strive to understand the “why” behind the behavior, not just the action itself.
- Derive Meaning — Ask, “What story is this behavior telling us?” rather than simply, “How many events did we log?”
- Design to Adapt — Human behavior changes over time, and your platforms must be flexible enough to change with it.
- Ground Everything in Ethics — Privacy and trust aren’t limitations; they are competitive advantages.
