All year, I’ve been having this dream.
I’m somewhere luminous—think an underwater utopia or a stained glass-grand library. Then, walls and floors shift, warp, and gloom into M.C. Escher staircases. I had the dream again this weekend, following a few days in San Francisco for Fortune’s Brainstorm AI conference.
Anxiety about a world in flux? Possibly. But I can’t help connecting it to the time I spend in my waking hours as a reporter covering AI, talking to people about a shimmering technology that only seems to move further away the more you focus on it.
In fact, this was a big theme during the Term Sheet Breakfast at Brainstorm AI, where several investors at the forefront of AI discussed the current state of the technology. "When it comes to our conception of the last mile in AI, we’re all going to realize it’s much bigger than we all thought it was going to be," Sara Ittelson, Accel partner, told the audience at the St. Regis last week. "It’s more like the last marathon, the last 50 or 100 miles, not the last mile."
Large incumbents hold an odd spot in this marketplace—ripe for disruption, but with a hazy expiration date. On one hand, enterprise adoption of AI is going to take far longer than perhaps many realized, Lightspeed partner Raviraj Jain said.
"I do think that getting AI to enterprises is extremely hard," said Jain. "Fundamentally, generative AI is generative, right? It’s not factual, it’s not performant. And, therefore, to build the core infrastructure to actually enable generative AI to make massive differences at the enterprise stage will take a lot longer."
On the other hand, change at the largest companies is still coming down the pike as the ways in which we work with AI shift. Legacy companies are increasingly vulnerable, said Tomasz Tunguz, Theory Ventures founder and general partner.
"If AI changes those workflows, there's an opportunity to sell the workflow software, and there’s an opportunity to unseat the existing systems of record," Tunguz told the audience, adding he believes $1.5 trillion in software spend that’s focused on the cloud is "now on the table."
Yet even as the ground keeps shifting beneath them, incumbents with large reach still hold substantial power. Cack Wilhelm, IVP general partner, cautioned that, if you’re a startup and you’re worried that Big Tech or OpenAI will release a feature rendering your company obsolete, you’re already losing: "If you’re relying on the hope that the next big model won’t subsume your product’s functionality, you’re probably doomed," she said.
Wilhelm’s point underscores the reality that, if you’re not doing something with a serious moat, weathering the next few years is unlikely. Simultaneously, there’s a sense that core changes that help AI reach its full potential are still to come. Steve Jang, Kindred Ventures founder and managing partner, describes today’s operating systems as "fundamentally broken" and believes there are changes coming soon that will affect how we interface with the world around us.
"If we have new operating systems that are AI-native, that understand inputs and outputs, you could have thousands of models flowing through," said Jang. "If you look at Netflix, it’s a system of models and video streams, right? That’s the beauty of Netflix, the recommendation engine and the personalization. So, I think AI operating systems will make that more possible for IoT, wearables, phones, computing, everything."
Both Jang and Wilhelm are investors in AI search unicorn Perplexity, which has made waves in the famously impenetrable search market (dominated for more than a decade by Google). These kinds of new operating systems could lead to a world that Perplexity has begun to demonstrate is possible—one in which AI changes our behavior on a personal level.
"I think it comes down to brand, and everyone's experience with it," said Wilhelm. "Allie and I were talking about how we're hourly active users, we’re not daily active users. If you build a habit and change consumer behavior, good things will happen." (Fortune is a member of the Perplexity Publishers’ Program, which shares ad revenue with publications when their content is used or referenced in a Perplexity answer to a sponsored question.)
One little-discussed area where AI could take hold is voice, a topic on which most panelists see a lot of potential. "We haven’t really explored voice as an interface since phones," said Kindred’s Jang. "If you think about all we’ve done with touch, video, and images as an interface, voice remains sort of elementary, and it's catching up now because of AI."
Rebecca Lynn, cofounder and general partner of Canvas Ventures, said the intersection of voice and AI is "the biggest change and maybe the biggest opportunity we don’t hear enough about." She believes that voice can open the doors to wide-ranging societal change, helping counter concerns like rising illiteracy rates. (More than 20% of U.S. adults were categorized as illiterate in 2024.) Voice AI can help those who struggle with reading better understand vital information, Lynn said.
The group of investors painted a picture of a more 'voice-forward' world, where AI has the potential to change our daily experiences, like the way we work in the office or visits to the dentist. But voice also represents what, perhaps, is the biggest unanswered question about AI. Where are the lines between tech and humanity, and how should the two co-exist?
"What happens when every agent you talk to is perfectly empathetic, perfectly patient? You’re never wrong. They never get mad at you," wondered Accel’s Ittelson. "What happens to people when they can engage with these AI personalities that don’t have the rough edges we all have as humans, that actually is the beautiful part of humanity?"
We don’t know how AI is going to change business because we don’t know how it will change us. And that’s quite an endless staircase.
See you tomorrow,
Allie Garfinkle
Twitter: @agarfinks
Email: alexandra.garfinkle@fortune.com
Submit a deal for the Term Sheet newsletter here.
Nina Ajemian curated the deals section of today’s newsletter. Subscribe here.
