Webinar: Real Talk on AI in AR. What’s Working, What’s Not, and What’s Just Hype.

What happens when you put three finance and technology leaders in a room and ask them, honestly, what’s working with AI in accounts receivable (AR) and what isn’t? You get a conversation that skips the hype and gets straight to practical reality. According to one study, most finance and accounting teams (78%) are already exploring or implementing AI. Yet the gap between exploring and actually deploying something that sticks remains wide, and the reasons why are consistent across every organization.

That’s what this webinar, hosted by the Controllers Council, set out to unpack. The panel brought together Greg D’Eon, VP Controller at LTK Commerce and the Controllers Council 2023 National Controller of the Year; Philip Peck, VP Finance Transformation at Peloton Consulting Group; and Ahsan Shah, SVP of AI and Analytics at Billtrust, the leading provider of AR automation software for B2B companies.

Here’s what the conversation found.

AI is Reshaping Finance Beyond Accounts Receivable

The panel opened with a broad scan of where AI is gaining real traction across the finance function. Peck set the big picture, walking through the full end-to-end process.

He called out automated journal entry classification, exception-based account reconciliation, and intercompany dispute detection as areas already showing strong benefits from AI investments. He also noted AI invoice capture and three-way cash application matching automation combining OCR and machine learning, as well as payment fraud monitoring. And in financial planning and analysis (FP&A) — the domain he described as a personal focus — predictive cash flow forecasting, rolling budgets with AI components, scenario modeling, and forecast bias detection.

“Something that spans different worlds is predictive cash forecasting,” said Peck. It’s an area that touches treasury, FP&A, and AR simultaneously, and one where the data demands of each group have historically made true convergence difficult.

What AI in AR Looks Like on the Ground

The session drilled into AR specifically, including the starting points, the workflows, the tools, and the realistic progression from manual to fully automated.

D’Eon described LTK’s approach to AI as a problem-first approach.

“The approach I took personally was: where are the biggest pain points in my team, and where are the biggest volumes of data we’re sorting through? We attacked cash first. We have 18 different bank accounts. The technology is now getting to the point where you can match a good chunk of that, whereas 3 to 4 years ago it was 50% at best.”

Greg D’Eon, VP Controller, LTK Commerce

Shah framed a key unifying theme as data synthesis. “When you think about accounting data, you’ve got financial data, you’ve got ERP and CRM data. Being able to scour through multiple systems and say there’s a drift here — that’s a huge win. And now we have the generative AI revolution, which opens the door to combining unstructured data like call notes and emails with structured financial data in tables and systems.”

D’Eon also described an AI agent LTK built that runs through overdue invoices, identifies the highest-value ones, and generates a daily follow-up list for the AR specialist to call. This agent replaces hours previously spent manually sorting through the urgent overdue invoices and those that can wait.

“We have an agent that runs through each of those invoices, pulls out the highest-value ones that will help us chop down that 120-plus bucket as much as possible. Those become the daily hit list for our AR specialist. A lot of times, we find an issue with payment application, so you have to go back and talk to the payment application specialist. There’s a lot of back and forth. But the AI does a good job of aggregating themes, so you can communicate to executives where you are with the state of collections.”

Greg D’Eon, VP Controller, LTK Commerce

Thinking Upstream: Preventing Problems Before They Reach Collections

Where LTK is focused on managing what’s already in the queue, Billtrust’s approach is to push the intelligence earlier. Shah described the goal as preventing invoices from ever needing a collections intervention in the first place.

“We’re thinking: how do you prevent something from getting to collections? Can I identify credit risk and credit behavior early? Can I optimize payment policies? We want the AR function to not be spending time calling and going through emails. Imagine automated phone calls, getting a transcript, and understanding the key issue. We want to remove that manual overhead entirely.”

Ahsan Shah, SVP AI & Analytics, Billtrust

Peck added that one of the most innovative AR deployments he’s seen is a client using AI to support contract-based billing and customer risk modeling as they expand into new markets, moving AI from operational support into a strategic growth tool.

The Biggest Mistake: Not Building Trust Before Handing Over Control

All three panelists agreed that the biggest mistake teams make is trying to deploy full automation before they’ve established trust in the system. Shah described the progression Billtrust designs into its deployments.

“Our customers don’t want to relinquish control to AI without gaining trust and having some transparency first. So, we spend a lot of cycles not just building AI solutions, but getting people acclimated. Can we start with basic insights? Can we tell you what’s wrong, and then move on to: Can I set this on autopilot? They gain the trust as we progressively go forward.”

Ahsan Shah, SVP AI & Analytics, Billtrust

The Metrics that Tell You Whether AI in AR is Working

Days Sales Outstanding (DSO) and the percentage of receivables past due by 120 days are the two standard reporting metrics that nearly every team tracks. But the panel pushed into more leading indicators, which tell you whether agentic AI deployments are performing before overdue invoices and delinquent payments show up in the balance sheet.

D’Eon outlined two metrics that LTK watches closely:

• Time-to-invoice after deals close, tracking what percentage of invoices go out within two days of a deal closing
• Time-to-cash-application, the gap between payment hitting the bank account and the invoice being marked paid.

“Those two,” he said, “tell you a lot about where the AR function is actually adding friction, or not.”

Peck added touchless processing rate as a direct outcome metric for automation, explaining what percentage of transactions go through end-to-end without manual intervention. On the customer-facing side, he also mentioned: invoice accuracy rates, dispute rates, portal adoption, and self-service utilization.

“These are innovative KPIs that measure how well you did in providing information to customers so they can actually make a payment,” he added.

Shah highlighted the value of debtor segmentation and benchmarking practices.

“We have data and signals across heavy equipment or manufacturing, for example, he explained. We know volatility is high in some industries, so we’re taking the view of: How did you do versus yourself? Are you on track? We get a lot of interest in industry-level benchmarking as a baseline to measure against,” he explained.

The Hardest Lessons: What Gets AI Deployments into Trouble

Every panelist had seen AI initiatives stall, and the failure patterns were consistent.

Data Quality is the Non-Negotiable Foundation

“The number one challenge is poor master data quality, and it goes hand in hand with another pattern. People don’t fix broken processes before they move to automation. You need to address the data and the process before you’re able to achieve the value of AI. Another problem is going for the brass ring before an organization is really ready. It’s easy to be too aggressive, moving too fast toward full automation. You have to have the foundational elements in place first.”

Philip Peck, VP Finance Transformation, Peloton Consulting Group

“AI is just going to amplify your existing current state. If your master data is not in a good place, AI is not going to solve that for you. But when you take the time and care — really thinking about your business process, understanding your data systems, getting that context layer right — it’s very, very powerful. And if you get the context layer wrong, it actually amplifies the problems and leads to going backwards.”

Ahsan Shah, SVP AI & Analytics, Billtrust

Find the Friction, Not the Flashiest Use Case

“The most practical approach to AI is this: What is my team arguing about the most? Where is the hot potato reconciliation that’s going from one person to another because nobody wants to do the legwork? That’s ripe for the picking. Find out how to use AI to solve friction within your teams, because it definitely can.”

Greg D’Eon, VP Controller, LTK Commerce

Peck also flagged change management resistance as a persistent and underestimated barrier, specifically the pockets within organizations that revert to old processes even after new solutions are deployed. “It’s critical to combine AI with workflow automation, keeping humans in the loop around governance and reviewing. Finance, IT, and business operations need to work hand in hand as you’re executing against these initiatives,” he said.

Will AI Eliminate Accounting Jobs?

The question came from the live audience, and it’s the one that comes up in every room where AI in AR conversations happen. The panel’s answers were consistent.

“My commitment to my team was: I’m not replacing anyone’s job with AI. I’m enhancing your ability to do more. If you take that approach, I think that’s probably the right way to view it.”

Greg D’Eon, VP Controller, LTK Commerce

“AI doesn’t replace jobs — it replaces tasks. Jobs evolve. People that know accounting, that know finance, that know their domain, are actually going to be very well amplified and positioned.”

Ahsan Shah, SVP AI & Analytics, Billtrust

Peck drew on his consulting experience to frame it as a change management opportunity. “If your team is stuck in the drudgery of low-value-add tasks, they don’t have the capacity to be indispensable business partners. Something we do with our clients is help paint a picture of how the world could be better for them at an individual level, at a group level, at a company and strategy level. Jobs change, probably faster than ever before. But it can be a vastly better place,” he added.

One Piece of Advice for Controllers and CFOs using AI in AR

The session closed with each panelist offering a single recommendation for finance leaders at any stage of their AI journey.

Greg D’Eon: “Look around your organization and identify where your biggest pain points are. Look at specific processes that are repeatable — the kind of thing that would go to an intern if they walked in the door. What can AI help us with? Don’t try to boil the ocean.”

Ahsan Shah: “Get in there yourself and test it. The more anybody in any department learns by using it day to day and making it part of their habit, the better your intuition becomes for where it can be applied and where it’s not that great. You don’t have to be super technical.”

Philip Peck: “Go through your existing processes and look under the iceberg. Find the long pole in the tent — the areas where if you make improvement it will drive substantial value. Start piloting. See the value in AI in AR. Get credibility in the organization. That steamroll effect builds on itself in a very productive way. Getting started is the number one thing.”

AI in AR: Three Takeaways from the Webinar

1. Start with friction, not ambition. The highest-ROI projects using AI in AR are the ones targeting the processes nobody wants to own. AI should absorb the most manual task — not the most sophisticated use cases.
2. Fix data before scaling AI. Poor data quality is the number one obstacle every panelist named, and research proves it. AI amplifies your current state — clean data multiplies results. Bad data multiplies problems.
3. Build trust progressively. Move from insights to recommendations to assisted actions to full autonomy. Each stage earns the right to move faster and automate more. Skipping stages is how organizations end up with AI their teams don’t trust and don’t use.

Want a step-by-step plan for digital transformation in AR? Don’t miss this blueprint for AI in AR.