Whenever I read a report from a prominent venture firm, I do so skeptically. First, their interests are firmly in hand—to attract and reassure investors and protect investments. Second, they have been stunningly wrong. For evidence, you have to go no further than May 2022, when Sequoia’s “Adapting to Endure” report predicted the 2022/2023 recession cataclysm. Which never materialized.
That report mystified those of us selling to the enterprise in 2022. I had the best pipeline ever in 2022. What I didn’t have was the backing of my investors. For non-existent reasons. That still smarts. It’s one of several chips I carry on my shoulders.
A March 21, 2024 report by Andreessen-Horowitz with the lengthy title, 16 Changes to the Way Enterprises Are Building and Buying Generative AI, sheds some light on how enterprises look at generative AI as part of their technology and product portfolios. I find this report compelling for startup founders and enterprise CIOs.
There’s gold in them there hills!
Chief Information Officers of large enterprises are under enormous pressure to incorporate AI into their technology portfolios. Some of that pressure comes from the top, CEOs afraid of being disintermediated or shown up by their competitors. A fair amount of consumer technology (e.g., ChatGPT in marketing) is entering the enterprise. CIOs have to get a handle on these homegrown efforts. A thousand flowers blooming can be a security and budget nightmare.
Though these leaders still have some reservations about deploying generative AI, they’re also nearly tripling their budgets, expanding the number of use cases that are deployed on smaller open-source models, and transitioning more workloads from early experimentation into production. “
CIOs are responding by dramatically increasing their spending in 2024. For the average enterprise interviewed for this report, spending on LLMs alone will increase by 2.5x this year from $7M to $18M
Savvy CIOs are also consolidating budgets. They raid innovation funds and departmental spending to fund mainstream IT investments. They’re also looking for funding in departments that have used AI to save operational costs.
One company cited saving ~$6 for each call served by their LLM-powered customer service—for a total of ~90% cost savings—as a reason to increase their investment in genAI eightfold.
Let’s simplify the language in that quote: Calls cost $60 to serve before LLMs, $6 now using LLMs, and $54 shifts to the CIO. It's not too bad for a CIO if he can wrestle the departmental budget from its VP.
Build vs Buy
Anyone who has been in enterprise AI for more than ten minutes has heard about Air Canada’s flub with a customer-facing LLM. This bad PR makes CIOs and corporate buyers reluctant to invest in AI, especially for customer-facing applications. So, most of the focus is on back-office applications.
I spoke to a board member of a large healthcare company yesterday. He mentioned that they looked at AI opportunities from three perspectives. Front Office. Back Office. Patient Outcomes. Patient Outcomes were off the table for now due to the uncertainty about risk. Front Office use cases were being evaluated cautiously. Back Office, saving money on billing and other operations with AI, was the most appealing and least risky. I think most businesses have used similar calculus.
The challenge for CIOs is the lack of supply. There aren’t many differentiated AI-powered technologies to help with those persistent back-office problems. Sure, most apps now have an “AI Button” [more on this in next week’s post], but those buttons are…just stupid. They’re LLMs wrapped around current tech. They do little to leverage AI's real power to resolve the business's relentless challenges.
With little real support from the vendor community, CIOs have taken matters into their own hands. Like the vendor community, they’re figuring out where to apply LLM wrapper technology. While they are experimenting, as illustrated by the increase in AI spending, they’re hungry for best-in-class AI tech from the vendor community.
While one leader noted that though they were building many use cases in house, they’re optimistic “there will be new tools coming up” and would prefer to “use the best out there.”
I believe this is the prevailing mood. While some CIOs fancy themselves intrapreneurs who are just as capable as any startup founder regarding product vision and software engineering, they are in the minority and delusional.
There are two lessons for entrepreneurs at this moment. First, help CIOs transition by selling AI picks and shovels. And second, don’t wrap an old problem with language. Use AI to solve an old problem in new ways using the transformative capabilities of AI.
Picks and Shovels
We believe that AI startups who 1) build for enterprises’ AI-centric strategic initiatives while anticipating their pain points, and 2) move from a services-heavy approach to building scalable products will capture this new wave of investment and carve out significant market share.
The lion’s share of AI investment today is in the pick-and-shovel business. OpenAI, Cohere, and Anthropic are AI platforms. They supply those who want to create AI technology. While software startups are a large part of their business, so are CIOs.
And while investors are loath to invest in services-heavy businesses because of their inability to scale like software, talent is the scarcest of resources. Any software deployed to help CIOs build will require services to help them implement. CIOs cannot attract the same talent level that startups and large cash-rich players like Google and Microsoft can afford.
Because it’s so difficult to get the right genAI talent in the enterprise, startups who offer tooling to make it easier to bring genAI development in house will likely see faster adoption.
Not only will services be critical, but automating some of the things that top-tier talent brings to the table will be of tremendous value to the CIO. I’m interested to see where Scott Wu’s Cognition Labs goes with his software engineering chatbot Devin. While Cognition’s work is currently focused on what I would refer to as traditional technologies, pointing it at AI will be the big bang for CIO gold mining efforts.
The Same Old, Same Old
While the pick-and-shovel business will do well for the near-term planning horizon, the real value will be in using AI in ways that fundamentally shift our ability to solve business problems.
we believe that the apps that innovate beyond the “LLM + UI” formula and significantly rethink the underlying workflows of enterprises or help enterprises better use their own proprietary data stand to perform especially well in this market.
This has always been true for startups. The greatest value is created and thus extracted by founders and investors by changing how things are done. Everything from machine learning to language models to knowledge graphs and semantic analysis, all “AI” technologies, can help with these pervasive problems.
Sure, language models are the revelatory technology of the moment. But they’ll only be part of revolutionary solutions. Entrepreneurs and the rare CIO with engineering chops will use a variety of current and emerging AI technologies to rethink business practices.
When AI investors and entrepreneurs shift to this mindset, and I’m already seeing it, they’ll return to their bread-and-butter of changing the world with extraordinary new ways of doing things.