Nikolai Braiden has spent his career at the forefront of the digital revolution, acting as an early adopter of blockchain and a seasoned advisor for startups looking to disrupt the status quo. As a resident FinTech expert, he has seen firsthand how technology can reshape the fundamental structures of digital payments and lending systems. Today, the conversation is shifting toward the rise of agentic AI—systems that don’t just suggest actions but execute them on behalf of human users. In this interview, we dive into the transition from traditional Know Your Customer (KYC) protocols to the emerging necessity of Know Your Agent (KYA), the escalating scale of synthetic fraud, and why data integrity remains the ultimate barrier to entry in a world run by digital proxies.
As digital agents begin managing financial transactions and scheduling on behalf of humans, the industry is moving toward a massive shift in how we handle identity. How does the concept of Know Your Agent redefine the way we perceive security?
The shift to Know Your Agent, or KYA, marks a fundamental pivot from verifying a physical person to verifying a digital machine that acts as their proxy. In this new landscape, we are moving beyond traditional KYC because we are no longer just looking for a human pulse; we are looking for the authority and data integrity behind a digital entity like ChatGPT or Microsoft Copilot. It is not about trusting the agent’s intelligence, but rather trusting the agent’s identity and the reliability of the information it is acting upon. We are now tasked with creating behavioral patterns and digital footprints for these machines just as we did for humans decades ago. To make this work, authoritative data must form the foundation for every secure, agent-driven interaction, ensuring that the “user” is exactly who they claim to represent.
Trust is often viewed as a human trait, yet we are now asking systems to trust machines. What are the specific technical and structural components required to establish a secure link between a human principal and their AI agent?
Trust in the age of agentic AI is built on a tripod of the entity represented, the scope of that agent’s authority, and the reliability of the data guiding its decisions. When an agent acts, there is a layer of abstraction that can obscure the end user, which means we must implement controls that verify both identity and intent simultaneously. This requires the industry to apply lessons learned during the pandemic, specifically using indirect signals and alternative forms of validation to confirm authenticity without physical interaction. We have to ensure the agent is linked to the correct human and is operating within a specific, authorized sandbox to prevent it from making unsupervised, high-stakes decisions. Ultimately, the limiting factor for any advanced technology is trust, and that trust is operationalized through the quality and integrity of the underlying data.
AI doesn’t necessarily create entirely new types of crime, but it certainly changes the velocity of existing threats. How is the rise of agentic systems magnifying risks like credential stuffing and synthetic fraud?
The real danger of AI agents is the sheer scale they bring to existing malicious behaviors, allowing bot attacks to try multiple combinations of identities and break into systems at an industrial volume. It has become significantly easier to conduct synthetic fraud, where a fake identity is paired with real business data to commit fraud on a massive scale without any human intervention. One of the most subtle risks we face is the loss of visibility; when an agent acts, a merchant might not be able to tell if an action is a legitimate automation, a system error, or a sophisticated malicious attack. Traditional security controls are not yet equipped to handle this level of abstraction, where the “user” could be a bot conducting identity probing at a speed no human could match. This amplification of risk means we have to rethink our defensive strategies to move as fast as the machines we are trying to stop.
You’ve emphasized that an agent is only as good as the data it consumes. What does a robust data due diligence process look like when onboarding data sources for these digital agents?
A robust due diligence process requires a deep dive into the entire lifecycle of the data, starting with how it was collected, how it is maintained, and how frequently it is updated for accuracy. At organizations like Data Zoo, we emphasize that before any data source is onboarded for identity verification, it must undergo a rigorous audit to ensure its integrity is beyond reproach. Because the data creates the agent’s persona, whatever flows into that decision engine will ultimately shape every financial or administrative choice the agent makes. If the data is siloed or constrained by legacy privacy issues, it limits our ability to verify the connection between the human and the machine. We have to treat data as the lifeblood of the agentic ecosystem, ensuring that every source is authoritative and verified before it ever influences a transaction.
Not every automated task carries the same weight, yet we often see a “one-size-fits-all” approach to security. How should organizations differentiate their controls based on the stakes of the agent’s decision?
We have to move away from legacy thinking and adopt a risk-based assessment where the level of verification matches the potential impact of the action. For example, if an AI agent like Microsoft Copilot mismanages a calendar and misses a meeting, the consequences are low and can be fixed with a simple reschedule. However, if an agent is tasked with a fund transfer or customer onboarding, the stakes are significantly higher, and we cannot afford to allow those actions without stringent, verified controls. Organizations must first conduct a comprehensive risk assessment of the specific services they are offering to determine where the high-stakes friction needs to be placed. By doing this, we can move away from siloed data and privacy constraints that have historically limited our ability to verify identities in the KYC space.
What is your forecast for the future of agentic AI in the financial sector?
I believe we are on the verge of a total overhaul of the data governance frameworks that have defined the last twenty years of FinTech. As AI agents become more autonomous, we will see a move toward more integrated, authoritative data ecosystems that break down the silos currently limiting our verification capabilities. The transition from KYC to KYA will become a standard industry practice, supported by new mechanisms that can verify the connection between a machine and its human principal in real-time. We will likely see a period of intense experimentation where behavioral patterns for bots are used to detect fraud before it can scale, essentially fighting AI with AI. Ultimately, the success of this era will depend entirely on our ability to build a foundation of trust rooted in the uncompromising quality and reliability of the data we feed into these systems.