Problem-First Thinking Is the Real AI Advantage

Most organizations enter AI conversations with a technology in mind. Michael Parlotto, Vice President of Emerging Technologies at InComm Payments, argues the sequence should run the other way. In this episode of Go Beyond the Connection, Michael shares the innovation framework his team at Go Studio uses to move from customer friction to measurable results — without leading with the tool.

Michael’s approach begins with problem framing: a structured method for uncovering the root cause of a customer challenge and ranking opportunities by business impact and execution difficulty. From there, the team moves into design thinking, building low-fidelity prototypes quickly to validate ideas with real users before committing significant resources. The final step is a futures practice that scans signals in the market, regulatory environment, and competitive landscape so launches stay protected long after they ship.

InComm Payments operates one of the largest embedded finance networks in the world, with more than 525,000 retail locations worldwide. That scale gives Michael’s team a ground-level view of where AI, voice interfaces, and resilient network infrastructure intersect with real customer experience.

Why Clarity Beats Code

“We use a framework called problem framing, and whether or not you know the problem, you can get to what the root problem is, and then you can prioritize that list of problems to get to what are the ones that are really the big bets that you’re going to use to make progress.” — Michael Parlotto

Michael explains that skipping problem framing is the most common — and most costly — mistake innovation teams make. Jumping to AI without first defining the customer problem means building something nobody uses. His team at Go Studio uses a yes-and workshop culture to pull in perspectives from across the organization, from engineering to finance to university partners, before any technical decision is made.

Data strategy follows the same principle. Michael’s research showed that data diversity consistently outperforms data volume when training AI models. A smaller, well-targeted dataset covering the range of real-world scenarios produces better outcomes than a massive but narrow one. Modern AI systems can also process unstructured data and handle imperfect inputs, reducing the barrier to starting.

Resilience as a Design Requirement

Embedded finance solutions and AI-powered applications only perform as well as the networks behind them. Michael walks through how his team designs for variable connectivity — building offline modes, edge processing capabilities, and intelligent synchronization so applications keep working whether a user is in a retail store, a healthcare facility, or a remote location with intermittent wireless service.

Security travels with the data at every hop. Michael connects network resilience directly to customer trust, particularly in healthcare environments where sensitive data moves across cloud, data center, and device layers simultaneously. This is where Bigleaf’s approach to reliable, policy-driven connectivity directly supports the kind of architecture Michael describes.

Episode Highlights

• How problem framing uncovers root causes and ranks the highest-value bets
• Using voice interfaces to improve AI input quality and customer intent clarity
• Why diverse data outperforms more data when training AI models
• How a futures framework protects launches from fast-moving competition
• Designing AI solutions that stay reliable on uneven and offline networks

Subscribe to Go Beyond the Connection to hear more conversations with technology leaders navigating the intersection of infrastructure, AI, and business outcomes. Find this episode on Captivate, YouTube, and all major podcast platforms.