Margaret and I drove through Hurricane Debby on Monday because of flight cancellations from Charlotte, and it reminded me of conversations with folks at the National Hurricane Center. While storms are extremely disruptive, the 20-30 mile wide eye of most hurricanes is relatively calm.
It is a good metaphor for what is happening in GenAI markets
Larry Dignan at Constellation analyzes recent earnings calls at Microsoft, Alphabet and others and summarizes what he calls a “halo narrative” around GenAI.
“Enterprise software vendors haven't been thrilled about the narrative that they aren't directly monetizing generative AI so the narrative is subtly being flipped” and
“GenAI is just software so don't go looking for direct revenue impacts.”
Larry and I had recorded a Burning Platform episode last November which presaged many of the trends which have led to this predicament and you can revisit them here
I would not scoff at any revenues, so don’t want to dismiss the halo effect explanation. However, to me it also highlights a lack of imagination at many vendors and the lack of appropriate data to create compelling GenAI use cases. I find Wall Street tends to often be myopic about CapEx spend at software and services firms, but in this case, we are throwing massive GPU and LLM resources at the problem, and it is not unreasonable to ask “where’s the beef?”
Developer productivity is showing up as a compelling GenAI use case for many vendors. They have plenty of previous written code in their own software, at their customers and at their partners to train their machines and at least generate a first draft of new code. SIs and outsourcers have plenty of data in 000s of proposals, custom code, test routines, data conversion, training modules and other assets and if they could get over their labor intensive mind set they could also deliver significant productivity with GenAI.
Vendors also have plenty of what Malcolm Frank calls SGA (sales, general and administrative) data to generate attractive (but not necessarily premium priced) GenAI use cases - watch this Burning Platform episode where we talked about raising the bar for AI. However, most vendors don’t have much COGS (Cost of Goods Sold) or Revenue or Balance Sheet data. They also don’t have many practitioners with depth in those areas. Operations and business models tend to be very different by industry and country and without that data and talent you cannot imagine or offer compelling AI use cases.
The chickens have finally come home to roost. I have said often that even after 25 years of SaaS applications, 75% of business processes across countries and industries do not have much vendor application choice. Vendors were content to keep selling basic ERP and CRM functionality, and there mostly for English speaking countries. If anything, the hype from GenAI caused a pause in expanding their transactional functionality.
Even SAP which has the broadest reach of customers, functionality and talent across industries and countries has been cautious in the reach of its GenAI offerings
The good news is the market is wide open for incumbent vendors and new entrants to build transactional functionality for so many unexplored markets. They can build their talent and data, and GenAI use cases for several niches. Many of these markets should be highly differentiated and command premiums which Wall Street expects to pay for NVIDIA GPUs and hyperscaler pricing.
There is one big caveat to all this. With the excitement around GenAI over the last couple of years, I have seen customers wake up to the realization that they themselves have the relevant practitioners and they also have the operational and revenue data. Why should they share that intellectual property with vendors and additionally need to pay for the privilege? Many customers are learning how to monetize that data. They will share such IP at very different price points.
The tech customers of yesterday could become tech vendors of tomorrow. Indeed that is a sub plot in my upcoming fiction thriller titled the ’The AI Analyst’ which is expected to be released by Thanksgiving.
Now that is fiction, but even in real life, data is the new oil and battles are already being fought where data is being stolen and traded
Gen AI is real. Finding use cases that justify the computing cost and talent are still elusive. It will take a new vendor mindset and business model to get use cases which justify pricing premiums that pay for the AI CapEx and practitioners.
But you need to get to the calm eye of the GenAI hurricane. The eyewall is way too turbulent. Even better get to terra firma. Driving through the relentless rain from Debby this week, we were glad we weren't flying through it.
