How are FIs adopting AI agents?

In this live episode of Good Question, Inscribe CTO and co-founder Conor Burke joins Jon Fry, CEO and founder of Lendflow, for a candid, behind-the-scenes look at how AI agents are already being used and tested by real-world fraud and risk teams.

Hosted by Brianna Valleskey, this interactive discussion covers practical AI use cases, live demos, and insights from a highly engaged audience of fraud fighters.

From semantic search and document analysis to the rise of fake, AI-generated documents, this conversation shows where agentic systems are making the biggest impact (and where caution is still warranted).

Four Agent Use Cases Gaining the Most Traction

Jon outlined four key areas where financial institutions are already seeing value from AI agents:

• Document categorization & extraction: Automating labeling and pulling structured data to reduce onboarding friction
• AI communications: Using agents to send document reminders, explain loan terms, or respond to missing info
• Data enrichment: Enhancing decision-making with cleaner, more complete data inputs
• Semantic search: Replacing filters and dashboards with natural language queries

These use cases share one common thread: they replace repetitive, manual work (not human judgment).

How Inscribe Agents Flag Fraud

Conor walked through live examples of Inscribe’s AI Risk Agent at work, including:

• X-Ray Detector: Surfacing hidden layers in PDFs where fraudsters have overlaid fake names or addresses
• AI-Generated Document Detector: Identifying documents created by image generation tools like Midjourney or OpenAI's DALL·E

While some generated fakes are easy to catch today, they’re rapidly evolving. “We expect this to be a much bigger issue six to twelve months from now,” Conor noted.

To stay ahead, Inscribe is investing heavily in tooling, training, and new types of forensic, network, and “agentic” detectors.

Inside the Agent Brain: How It Scores and Summarizes Risk

Audience members asked some of the sharpest questions of the session, including how Inscribe’s AI agent weighs risk and explains its reasoning.

Conor shared that agents use a mix of LLMs and classic models (like logistic regression) to assess fraud signals, then summarize their findings in plain language with supporting documentation and next-step recommendations.

These summaries:

• Provide transparency for audit and compliance
• Give frontline teams more confidence in agent decisions
• Help bridge AI analysis with human oversight

You Can Talk to Your Agent (Soon)

A crowd favorite moment came when Conor demoed Inscribe’s newest feature: natural language fraud search.

Instead of SQL or dashboards, teams can now ask questions like:

• “How many AI-generated docs have we received this month?”
• “What’s the most common fraud signal we’re seeing this quarter?”

And get clear answers, in seconds.

Jon called this a major unlock for adoption: “We’ll soon be talking to agents the way we talk to our teams today. That shift will happen faster than people think.”

Trust, Transparency, and Human Oversight

The conversation ended with a powerful reminder: AI agents aren’t replacing analysts. They’re augmenting them.

Both Conor and Jon emphasized the importance of:

• Bias testing and continuous eval loops
  
  
• Internal risk ops teams to challenge and refine outputs
  
  
• Partnering with vendors who build with responsibility in mind

“You don’t just flip a switch,” Bri added. “You test, you measure, and you expand from there.”

Final Thoughts

“2025 might really be the year of the agent,” Jon said.

“We’re just scratching the surface of what these systems can do,” Conor added.

But one thing’s clear: AI agents are no longer theoretical. They’re already shaping the future of fraud prevention.