Over the years we have done a lot of CFO work for companies. One of the key things we have learned from that experience is that CFOs needs to work very hard to avoid becoming Dr. No, Too often, CFOs fall into the role of the person who says no to everything. That being said, we have to think a big wave of role playing is coming down the pike this year or next as the costs of AI start to pile up.
Judging from the vendors’ marketing materials and the general media, every company in the world is racing to find an “AI strategy”. Obviously those are not the most reliable sources for this kind of information, but we do think it is safe to say that big companies everywhere are having to feel about in the dark to understand what AI means for their business. Over the years, we have seen many hot technologies come and go, leaving perplexed corporate buying teams reconsidering their purchase priorities and life choices.
We suspect that many will be tempted to file AI as just one more flavor of the moment and move on – send us an update when the computers come alive. But that would be a mistake. An understandable one, but a mistake nonetheless.
As we noted a few weeks back, in its simplest form AI – really machine learning based on transformer neural network models is going to deliver some form of “under the hood” improvements in compute performance. Small tweaks which can have meaningful impact on costs and productivity.
However, one of the dangers of the massive hype aura flowing around AI today is that it raises expectations to unsustainable levels.
Imagine Acme Box Company, proudly producing the best in corrugated containment solutions since 1890. The IT team there has been reading up on AI. Like everyone else they have gotten excited about the prospects of these systems and seen the enthusiasm, and high salaries, that their early AI adopting peers enjoy. They want their company to build an AI model to improve Acme’s systems and bring their proposal to their CFO. They have done their homework, they know there is no way they can build a giant model, but they think they can build a smaller model trained on their ERP, HR and CRM systems data. They will need to rent a lot of GPU space from their cloud provider to train all that data, and this will cost several million dollars.(They thought about buying their own GPUs, but changed course when they saw how much those cost.)
The CFO saw this coming, she is very good at her job, and so has done her own research. Despite that, she is still a bit shocked at the amount of money the team is seeking. It would pretty much evaporate the corporate bonus pool. Will this new system allow us to automate invoicing or onboarding and training new employees or lead to a boost in sales? Will it allow them to stop licensing any of their major, very expensive incumbent software platforms? What, in short, will it do?
Here the team has to pause. It’s not that they hadn’t thought about these questions, it’s just they were so excited by the prospects of the technology, and the boost to their resumes, that they thought everyone else would be excited about it too. The title of their presentation is “Embrace the Future”. They tell her that no, the system will not be able to do any of those things she asked for. They do think it will allow them to reduce down time on their equipment by 9%, it will reduce invoicing errors by 13%. It will shorten the customer service feedback loop by weeks. And it will allow their quality assurance team to be 6.5% more efficient. But they also admit that it will take six months to implement the findings from the model, as it will require re-plumbing some of their internal systems to accept inputs from the AI system.
A good CFO will take a look at those figures and know fairly quickly if they are sufficient to justify the investment. A reduction in equipment downtime alone could pay for itself very quickly. If it works. On the other hand, most people in that role will have already built up a fair amount of AI fatigue. The other CFOs at the Country Club talk about AI all the time, even though none of them can point to any real gains. Investors have been asking about it non-stop for a year, and seem to have some very unrealistic expectations of what it means for a box maker. A perfectly reasonable person could come to the conclusion that it is still too early, AI is not yet mature enough, to merit a multi-million dollar investment right now.
In our view, the big downside of all the widespread attention that AI garners right now is that it creates an easy scenario for companies to say no. Many people in these roles will remember the Internet Bubble. Much of the promise of those dot.com companies came to fruition, but only after a decade of learning and adaptation. It is very possible that some form of similar pattern recognition is kicking in now among the non-technical decision makers in the corporate world. That may not be the right choice now for some, but for many it might be.
Photo by Google Gemini
Leave a Reply