There is a push for CFOs and finance teams to understand and implement AI technologies. This is no secret. But there is also hesitancy, which a variety of data points tend to back up, that finance teams are not ready to jump in with both feet. While the profession is ripe for disruption, it does feel a bit like we’re seeing a shadow of the “what will a post-Excel world look like?” vibe.
I’m attending this year’s Gartner CFO conference, and the organization has an obvious interest in enabling companies to push through impediments to implement and apply AI technology. (Let me know if you’re in attendance; we always enjoy meeting our readership in person).
During the opening keynote, titled “AI Stalls,” the speakers addressed the barriers that may be holding CFOs back from implementation. The one that surprised me is the same factor that is arguably at present the biggest seller for AI technology.
***
That is, managing costs, and the threat of cost overrun. One of the technology’s biggest advantages is to be able to do things better, faster and cheaper because it frees up finance talent to do other things to drive the business forward. But while the tech’s promissory note of sorts is to bend this operational cost downward over time, the more a company uses it, the opposite may be true.
Keynoters Nisha Bhandare and Clement Christiansen discussed three types of cost overrun threats: the rollout, the ongoing and the experimentation.
Rollout costs for AI adoption include the infrastructure, talent acquisition and implementation into operations. This type of cost is nothing new to finance chiefs.
However, “there’s a uniqueness to AI costs. Given how new AI is, CFOs don’t really know how much it costs: they are learning as they go, driving cost estimates off by 500-1000%,” Bhandare said.
Ongoing costs include the cost of maintaining the large language models (LLMs), compliance, ESG considerations and the rate at which an increasing number of employees increase their rate of queries. The closest analogy I can think of to this is not a SaaS licensing model, but a legal research software model where companies such as Westlaw and LexisNexis would not only charge a licensing fee, but a querying fee as well in certain situations that could rapidly spiral out of control.
Additionally, Gartner highlighted experimentation costs as “sunk costs,” which seems to be a bit of a contradiction because sunk costs are normally related to a business’s costs that cannot be recovered. But in this case, the experimentation is necessary, but also ongoing, and therefore would bend the cost curve upward at a nonlinear rate as more and more groups within the enterprise adopt.
These three cost buckets combined mean the cost of AI will increase the more you use it, but in ways you may not easily anticipate or be able to forecast. To make financial sense, the upward bending cost curve therefore has to theoretically intersect, at some point, with the hypothetical downward cost of scaling efficiencies through operations over time.
The keynoters said, based on their own surveys, that 73% of finance leaders will increase technology spend over time. The spend makes sense, but CFOs — especially those still concerned with inflationary forces and a stagnating economy — will need to ascertain how long it will take the savings to kick in before the rising ongoing and experimentation costs eat up their budgets.