Andrew Moore, former head of Google Cloud AI, says building investigative AI agents for risk management workflows is like “solving a huge game of Sudoku, with literally millions of little facts, trying to find a pattern in it all.” 

Moore’s fintech startup, Lovelace, which exited stealth mode in late April, is building investigative AI agents that analyze vast data sets to help institutions fight financial crime, monitor and predict market changes, identify risk and bolster compliance operations, Moore told FinAi News.  

Moore’s 14 years at Google helped shape his strategy for Lovelace through learning to maximize extensive data and apply advanced AI to tangible use cases, Moore said.

“What I learned during that period was all the ins and outs of what it takes to do this translation of very theoretical, advanced research coming out of these brilliant, incredibly expensive research and engineering labs, and determining how to actually help systems run more efficiently,” he said.

For example, Moore’s experience overseeing fraud prevention at Google translates to anti-money laundering initiatives in banking, he said. 

“Anti-money laundering investigations inevitably involve lots and lots of data points,” he said. “You can’t just look at one transaction. … It’s how that transaction fits in a massive pattern and things that change over time, whether that’s geography, different financial instruments, different industries and so forth.”

Lovelace has raised an undisclosed amount in a seed funding round led by RRE Ventures, with other investors including Magarac Venture Partners, United States Innovative Technology and Carnegie Mellon University, according to a Lovelace spokesperson. Moore cannot yet name its banking partners due to non-disclosure agreements, but said early adopters comprise large and medium-sized institutions in the United States and United Kingdom. 

Context engine 

While data quality is crucial for effective AI agents, successful deployment starts with extensive model training to establish a “knowledge graph” or “context engine,” especially when dealing with millions of data points, Moore said.  

“When the agent starts out with its question, before it ever has to touch all that data in the background, you can formulate a really good plan for what data the agent is going to work with,” Moore said.  

These knowledge graphs also help ensure explainability and auditability, which are a priority for FIs when vetting potential fintech partners, he said.  

“You have to have multiple independent lines of reasoning,” he said. “When [agents] finish their work, they’ve always got to show this unbroken chain of bits of information they used as evidence. … You have to actually show a high-quality statistical analysis, as well.” 

Leaving tech giants 

Moore joins others who have left tech giants to start fintechs, including:

• Carmelle Cadet, left IBM after 10 years to start central banking infrastructure provider Emtech;
• Twitter co-founder Jack Dorsey, left to start payments processing and merchant services platform Block, formerly called Square; and
• David Marcus, left Meta to start payments company Lightspark.  

Register here for the FinAi Lending Summit, set for Oct. 7-8 in Las Vegas.