From A Simple Question to Agentic AI: Rethinking Customer Needs in RegTech

A recruiter from a RegTech company once asked me a simple question during an interview: “How do you know the customer actually needs this feature from your software?”

At first, it sounded straightforward. But the more I thought about it, the deeper the question became. It wasn’t just about product design or user experience. It was about something more fundamental: understanding real problems before building solutions.

Looking back, that question taught me two important lessons: first, how crucial it is to understand the real problems customers face, and second, how emerging technologies like agentic AI are beginning to address those problems in fields like AML (Anti-Money Laundering), CTF (Counter-Terrorist Financing), and KYC (Know Your Customer).

I/ The Question Behind Good Software

In many technical projects, especially when we’re starting out, we focus on building. We think about architecture, frameworks, performance, or adding more features. But this question pushed me to think differently: How do you know the customer actually needs this?

In RegTech, the “customer” is rarely just one person. It can include:

• compliance officers reviewing suspicious transactions
• analysts investigating alerts
• risk teams monitoring customer profiles
• regulators requiring transparency and auditability

Understanding their needs isn’t just about asking “What feature do you want?”. In fact, experienced product teams know that customers often describe symptoms, not the underlying problem.

That’s why many companies rely on structured discovery techniques, like SPIN selling, which focuses on understanding the customer’s situation and challenges before proposing a solution.

In practice, this often comes down to asking simple but powerful questions:

• What problem are you trying to solve? This helps identify the real objective behind the request.
• What does your current process look like? Understanding the existing workflow reveals where software can actually add value.
• What happens when it breaks down? Failures often expose the most critical pain points.

In fields like AML and CTF, these questions are particularly important because compliance workflows are complex and highly regulated.

For example, imagine building an advanced machine learning model to detect suspicious transactions. From a technical perspective, the model might perform extremely well. But if compliance analysts cannot understand why an alert was triggered, the system becomes difficult to trust or use.

That’s why successful AML software often prioritizes:

• explainability
• clear investigation workflows
• efficient alert management

In other words, the best systems are not only technically advanced, they are designed around real operational needs.

That question made me realize that building financial technology isn’t just about algorithms. It’s about understanding how people actually work, where the process fails, and how software can genuinely improve it.

II/ How Agentic AI Is Changing AML, CTF, and KYC

Around the same time I was thinking about that question, another shift was happening in the industry: the rise of agentic AI.

Unlike traditional AI systems that respond to individual prompts, agentic AI systems operate with goals, planning, and multi-step reasoning. They can break down tasks, use tools, and iterate toward an objective. In compliance, this is a paradigm shift from “alert me when something’s wrong” to “handle it and tell me what you did.”

From Static Rules to Intelligent Investigation

Historically, AML systems relied heavily on rule-based transaction monitoring:

• Transactions above certain thresholds
• Activity involving high-risk jurisdictions
• Rapid movement of funds between accounts

These rules work for known patterns but generate large volumes of alerts, many of which turn out to be false positives. Agentic AI introduces a different approach. Instead of simply triggering alerts, AI systems can assist in investigation workflows. For example, an AI agent could:

1. Analyze a suspicious transaction
2. Retrieve related customer information
3. Examine transaction history
4. Identify connections with other accounts
5. Generate a structured explanation for investigators

Rather than replacing analysts, the AI acts as an investigation assistant, helping them understand complex patterns more quickly.

Smarter KYC and Risk Analysis

KYC processes also involve large amounts of data: identity documents, transaction behavior, sanctions lists, and corporate ownership structures. Agentic AI can help by:

• Navigating multiple data sources
• Extracting relevant risk indicators
• Updating customer risk profiles dynamically

Instead of static onboarding checks, the system becomes continuously aware of risk changes. For compliance teams, this means moving from periodic reviews toward ongoing risk monitoring.

Why RegTech Companies Are Interested

For RegTech companies, agentic AI is attractive because it addresses several long-standing challenges:

• Reducing alert overload: AI agents can pre-analyze alerts and highlight the most relevant signals.
• Supporting investigators: Instead of presenting raw data, AI can produce structured summaries of suspicious activity.
• Improving explainability: Agentic systems can document their reasoning steps, which is crucial in regulated environments.
• Automating repetitive tasks: Data gathering and cross-referencing across systems can be handled automatically.

In essence, the goal is not full automation but better decision support.

Coming Back to the First Question

Looking back, the recruiter’s question makes even more sense today.

Technology in RegTech is evolving quickly , from rule engines to machine learning, and now toward agentic AI systems. But the core challenge remains the same: Does the technology actually help the people who use it?

Compliance teams don’t need more complexity. They need systems that:

• surface the right signals
• explain risks clearly
• and integrate smoothly into their workflows.

In that sense, agentic AI has real potential. Not because it is a new technology, but because it can adapt to the investigative nature of compliance work.

And maybe that’s the real answer to the question I was asked in that interview: You know customers need something when the technology makes their decisions easier, clearer, and more reliable.