Scaling AI: Is this the year it happens?

“This year is really about making AI operational and putting it into practice. There’s been enough evidence of success now, and enough people experiencing it in real life, where they know it’s real and it will have a real impact.”

So says Rob Israch, president of Tipalti, who has “pretty much been working with CFOs” his entire career. After a seven-year run at NetSuite, he’s spent nearly 12 years at Tipalti, during which time the AI-powered finance automation platform has grown from around 25 employees to more than 1,000.

Israch spoke to CFO Brew about where AI is being put into practice and how companies that are struggling with AI adoption can scale AI projects beyond the trial stage.

This interview has been edited for clarity and length.

Some companies are struggling to advance AI beyond the pilot stage. What are the barriers at this phase?

We’ve done a lot of CABs, or customer advisory boards, in finance, and what they’re coming back saying is, one, “The first thing is we really have to trust the numbers.” And they know AI is not perfect, so whatever is deployed has to have an oversight, it has to have transparency, it has to have auditability. They have to understand why it came to its conclusion, and it has to have the person in the loop to approve it and review it.

The other piece is it needs to be interoperable with all systems and data you have in place…Finance generally would rather look to their current systems and make sure their current vendors have an AI stance and a track record and a strategy and lean on that more heavily versus buying yet another system and trying to integrate it or buying an AI generic tool and integrate it in there.

What led to the development of the agents in your products? Was it something customers were asking for?

We’ve always had machine learning in the product but several years ago we were eyeing agentic AI already because you could see it around the corner…A lot of these agents are use cases that are prominent to our customers, where they’re like, “I would like to see this even better and even more automated,” or, “This is problematic for my day-to-day work,” and the agents are very much focused on those time-consuming, problematic, error-prone parts of their processes.

Duplicate detection is a great example of that, right? It’s a pretty bad thing to pay a supplier twice or more, a lot of money for things that they didn’t do. And a lot of times you don’t even know how to detect that…One of our agents identifies that, and it’s already identifying pretty large savings on a monthly basis for our clients by proactively identifying those duplicates before they happen and flagging them.

What advice would you give CFOs about scaling AI?

It does start with the top. Chen [Amit], in particular, our co-founder and CEO, he’s been great at championing it, helping us understand what the potential of AI is…And then of course, making sure all the C-suite are on board and finding champions in the business…Those champions probably will have minimal impact, unless you have the CEO on down pushing it down and getting everyone onboard.

Could you give me an example of one of the AI tools you’ve built in-house?

We have an agent that every employee in the go-to-market and the customer-facing group can pretty much ask any question of—anything about our solution, anything about the landscape positioning, messaging—and get an answer surfaced. Within five seconds, generally more like three, it answers a question. It’s pretty darn accurate on message, on brand. There’s obviously room for improvement on everything, so we’re tweaking it. But it’s not static; it’s not just, “Here’s a deck you can look at”…That’s the type of stuff that 10 years ago, you could not imagine—five years ago, even.