I don’t envy the CFOs out there right now.
AI-related costs are exploding. In many cases it’s difficult to see the ROI. Or at least the immediate ROI.
We spoke to a midsize company recently that is spending about $250k a year on ChatGPT licenses. Most of that usage is generated by a handful of super users. Leadership is weighing a massive org-wide training push to get everybody else up to speed.
If I’m the CFO over there, I’m like, man, we’re already spending $250k a year on licenses and we still haven’t onboarded most folks. Most folks will get in there and save time. Be more productive. But what is that going to do to spend? How does that map to return for the business? And how fast should it turn into ROI?
I think most of the CFOs I’ve met in my life care about two things: does an investment save costs, or does it drive revenue.
One of the most interesting areas of AI over the next couple of months, quarters, or years will be mapping AI investment to the business outcome.
It’s hard enough when everybody sits in one foundation model. Claude and ChatGPT offer credible enterprise dashboards that can get you started. But when you start layering in wrapper products and purpose-built AI tools, aggregating, mapping, and allocating token spend across teams, initiatives, and returns gets really tricky.
But.
The best CFOs I’ve known also know that not everything needs to show immediate ROI. New technology needs some time and space for experimentation.
Back to that midsize company. It’s hard to argue against onboarding all these employees and getting them up to speed. Individual productivity will skyrocket, even if they are probably going to burn tokens on things that are not immediately ROI-driven. What’s the value to your org of having a company that’s skilled with the new tools? How much experimentation do you allow for? How do you budget for that? When you get to steady state, how do you allocate tokens and map them to performance?
These are hard questions.
The team and I at Legible are hosting a roundtable discussion tomorrow to dip into some of these questions. Join us if you can, event page is here.