Private equity, middle market CFOs must push through AI’s ‘messy middle’ deployment

Sponsored by

Ready to move past the Wild West of AI experimentation? Try this decision-making framework that PwC’s experts came up with. It outlines where to lead, lag, or exit on AI strategy. Get the details.

Like any good coach, a CFO should have a well-thought-out playbook for their company. But when it comes to AI playbooks, some CFOs, both at private equity and in portfolio companies, are finding it difficult to cross their Xs and finish their Os.

According to a recent Accordion survey titled “The PE AI Adoption Benchmark: From scaling to systematizing”, many firms are stuck in “pilot purgatory”: They’ve spent capital on AI experiments, “but [there’s] no operational playbook to turn isolated deployments into compounding portfolio value,” according to the report.

But Kyle Roemer, managing director and US head of data, analytics, and AI at Accordion, said prototyping an AI use case is one thing; it’s entirely different to “get it deployed into your system…and into everyone’s workflow, so they can start leveraging it.”

Put more simply, the report states: “The majority [of PE firms] are somewhere in the expensive middle: past casual experimentation but not yet at genuine operating leverage.”

But portfolio company CFOs are “in a unique position” to execute on PE’s AI deployment plans, Roemer also said, given their exposure across “value creation areas of the business,” and connection with the PE sponsor.

And now that AI can produce more use cases in areas like P&L and forecasting, Roemer said it could give accounting teams more time to spend on strategic work, and companies an even higher exit value.

But Shawn Cole, president and founding partner of Cowen Partners Executive Search, emphasized that a CFO’s AI deployment needs to follow a clear value plan and path.

“It’s like R&D, and if you don’t have a consultancy that knows what to do, or how you’re going to benefit from it or an existing software, you’re kind of out in the cold,” Cole added.

Squeaky clean. The basis for deploying AI applications across an enterprise is operating on clean data, he said, meaning it isn’t duplicative, it has the same definition, and it can scale with the business.

Roemer—who’s spent nearly 20 years connecting data, analytics, and value creation for a variety of enterprises—defined clean data as “data coming out of your source system—your source systems could be a financial system, a CRM, could be a variety of tools—[that] represents how the business operates.”

“It’s one thing to have clean data and fix it at once; it’s another thing to make sure it’s governed properly, so that when you continue to grow at the scale you’re growing at, don’t introduce new quality issues to it,” he added.

Value added. Nearly nine in 10 operating partners Accordion surveyed—86%—said they “expect buyers to pay a premium for AI-enabled finance capability within two years.” Roemer said “forecasting is a great place to start” implementing AI.

“Where LLMs, in particular, are quite good is producing first passes at these things, so if you don’t have a really robust revenue forecast today, you can take a first pass with Claude or OpenAI…to get that version to then iterate on,” he added.

Private equity firms’ craze over AI-native and integrated companies is nothing new, but companies investing heavily in AI integration beyond the “pilot purgatory” can mean the difference between a deal and no deal, Roemer said.

“Especially in the middle market, most of the systems are messy, the data environment is messy. And if you get an asset or look at an asset that’s invest[ed] the time to automate a bunch of the back office, to introduce more leading technologies that don’t require as heavy a lift from a human capital perspective, it’s been true for years now: You pay a little bit more for that, because you can scale that with much more confidence, and without that infrastructure investment,” he added.

Cole added that where real value can happen is when AI eventually moves beyond just a plug-in to other financial software systems, and AI applications can operate independently and communicate cross-platform to deliver solutions.

“AI is truly in its infancy…when this becomes mainstream, it’ll be when the large accounting software systems, and forecasting, the Power BIs of the world, all chat with each other and give this to you as part of the solution, and not some add-on,” he added.

“This is like Microsoft Word, then there’s this plug-in that kind of works, but you’ve got to know how to use it, and it doesn’t really work that great, and you really need Microsoft Word to just make it a part of the system.”

Take a (short) breath. Roemer said that a lot of individuals on FP&A and accounting teams are tied up “doing manual data pulls, combining Excel files, doing a lot of, frankly, mundane work.” AI, of course, can free up their time by automating certain tasks, but Roemer said he thinks it’s still vital for people to be involved at the end of the AI forecasting process.

“People putting these revenue forecasts together are one thing; people then interpreting and figuring out ‘Why is something actually happening?’ and putting in the thought, the analysis, after the forecast is built, no one has time to do that…By [using AI to forecast], that allows them to be more strategic partners,” Roemer added.

But the idea that AI will suddenly give these workers, and their CFOs, time back in their day? Roemer calls it a misconception.

“With all the efficiency, I find even with myself, I’m just working more, I’m doing more. I’m more efficient, but I’m just working more; I’m getting to things that I historically could not have gotten to, [and] I think that’s really important,” he said.