PEX CFO Luke Pritchett is looking to build a “shadow ledger” for his company with the help of artificial intelligence.

The aim isn’t to replace PEX’s existing processes, but to create a “parallel version of our accounting ledger” that can be run autonomously by AI agents and sit alongside the company’s books as another way Pritchett as finance chief can check for mistakes, he told CFO Dive in an interview. The hope is that the shadow ledger gets to between 80% or 90% complete by the end of this year, he said.

Founded in 2006, PEX offers spend management solutions including corporate cards, receipt management and spend controls.

Starting the AI journey

Pritchett’s personal AI journey began early this year, when he tapped the tool to create an agent to help manage his children’s schedules, he said.

“That was really my first experience with using AI models, and it changed how I thought about my job at PEX,” he said.   

He first joined PEX in 2015 as a senior vice president, and took the CFO seat in 2022, according to his LinkedIn profile. Before the New York-based spend management company, he spent seven years at card network Visa in roles such as senior account executive for its prepaid partnership development team.

During his past decade at the company, he’s helped run the finance department from its early stages to its current level serving thousands of businesses. The company is also moving to make more use of AI.

In July, the company raised $160 million in debt and equity funding led by Bluff Point Associates. It aims to channel the funding to three places: expanding its charge card program, investing in employees and a “serious investment in AI,” according to a company blog post at the time.

When Pritchett, an AI enthusiast, considers the technology, “what's crazy is how much my job has changed” over the last six months compared to the previous nine years, he said. Following his success in his personal life using the technology, he moved to utilize it to solve for challenges or snags in the financial workflows at PEX, he said.

The first AI agent he created was an expense policy agent for internal use, he said. Members of the company’s finance team spent a lot of time seeking out missing information or receipts for employee expense reports, and the agent “really handles that follow-up loop,” he said.

“That way, when the month ends, most of the information is collected from the workforce, so the back office can focus wrapping up the remaining month-end items,” Pritchett said. The company also recently wound up rolling out one of the tools attached to that agent — a lost receipt affidavit that can be uploaded if needed rather than having users block the attached card — to its customers, he said.

As well as freeing up time for the finance team — who used to have to manually enforce the expense policy — employees themselves can query the agent on Slack about the policy to gain a better understanding of it and what they need to provide, he said.

“I think that might be the most valuable thing: not necessarily telling an employee that a transaction was out of policy, but helping train and educate employees on actually what our policy is in an easy environment that they can do from their phone or Slack their desktop,” Pritchett said.

Building connections

The expense policy agent itself is part of Pritchett’s larger project of building out a more autonomous, AI-powered back office, he said, using the technology to help share information between different systems. Pritchett also began tapping AI to build out connections between the company’s PEX or corporate card platform, its HR system, bank feeds and accounting systems using AI agents supported by Anthropic’s Claude through the company’s enterprise license, he said.

Those agents then “move data points between those systems of record, and then I have everything the agents do displayed on a dashboard, so I can track it just to provide visibility and accuracy and making sure it's creating value,” Pritchett said.

Pritchett has created usernames and passwords so members of the finance team, such as PEX’s controller, can also access the AI dashboard, he said. The team is slowly becoming more engaged, especially as more systems are integrated, he said.

As CFO, Pritchett aims to encourage more experimentation with AI in a way that is collaborative, he said. For example, if individual employees each tap generative AI solutions, what Pritchett’s observed is “you get roughly, 20%, 30% more done,” he said. “That's great productivity. It's worth having. It's fantastic, but it's a little bit linear and it caps out because the work stays trapped where it's being done.”

He is currently brainstorming on how to move beyond that; if each employee “does have a specific agent, there could be more agent-to-agent collaboration,” he said. “I think that's when you can break the next ceiling of productivity and excitement.”