VCs sit through a lot of pitches about AI every day.
Too often, it’s the same handful of ideas presented on pitch decks bearing slightly different fonts, illustrations, and numbers.
While their eyes glaze over, many of these VCs are secretly wishing someone would come along and pitch them something else entirely; something for that big market opportunity, demographic, or problem that everyone else is ignoring.
If only these VCs could speak up and say what they really want. Coins2Day is here to help!
We surveyed nine VCs about where we’re at in the AI boom, and what they’re keeping an eagle-eye out for. Here are the nine AI startup ideas that VCs wish someone would pitch them on.
1: Reinventing recruiting
Recruiting is in desperate need of an AI glow-up, says NFX founding partner James Currier.
“I’ve been waiting this whole time for someone to reinvent recruiting on the B2B side,” Currier said. “We have examples of people incrementally saying ‘I’m going to do this to resumes, or I’m going to do this for the applicant tracking systems (ATS). These ATS systems never got that big. There are tons of them, none of the businesses ever got that dominant.”
Numerous recruiting companies have come and gone since the early 2000s, trying to solve this exact problem, but few have endured.
“Recruiting has always been a mess,” Currier adds. “But I think with AI, you could just reinvent the experience. It’s been over a year-and-a-half, and no one’s done it. Let’s just throw out everything we know about it. It’s such an amorphous problem, it’s perfect for AI.”
2: Materials Discovery
For Tom Biegala and Ben Hemani, founding partners at Bison Ventures, they want to see more companies at the intersection of AI and materials discovery––using artificial intelligence and other advanced computational approaches to develop new materials with novel properties and applications. Think new more biodegradable and reusable plastics, new alloys that are lighter weight and cheaper to make, and new materials for batteries.
“AI has been talked about for a very long time as a way to accelerate materials discovery,” says Biegala, Founding Partner at Bison Ventures. “I have not seen a company do that well. So far, I continue looking at businesses. It’s just a very hard nut to crack, and you need a very unique kind of combination of AI expertise, plus material slash chemical expertise, plus high throughput experimentation, plus amazing engineers that understand the ins and outs of these various use cases and how to bring materials to markets.”
But this one in particular would take a very diversified founder team, with expertise in AI, and chemicals, engineering, and materials.
“AI capabilities are really counterintuitive for people who are not researchers in the space,” says Hemani. “So to use AI to move the ball forward on novel material discovery is very non-obvious and requires really great expertise and very specialized applications of these tools. It also can’t just be a black box.”
3: An AI speed reader
“The miracle of AI is not that it writes for us, but that it reads,” says NFX’s Currier.
“So when you think about looking for a job, you’re reading through a lot of job posts. And, when you’re a recruiter, you’re reading through lots of resumes and lots of cover letters. “I think on the consumer side, there’s an opportunity for travel, real estate, dating, and then credit cards and other financial products… When you’re looking for a trip, when you’re looking for a date, when you’re looking for a home, you’re reading, reading, reading.”
“This thing reads better than we do,” Currier adds.
It’s also possible that AI will tell us what to read soon enough.
“AI to manage my information consumption: set me up with books, podcasts, and articles based on my interests, and help me continue to learn and absorb information in a seamless way across platforms, including when I’m commuting and on the go,” says Mike Ghaffary, General Partner at Canvas Ventures, via email.
4: An open-source Nvidia CUDA alternative
Nagraj Kashyap, General Partner at Touring Capital, has a really specific idea: Build an open-source alternative to NVIDIA’s CUDA software that would allow AI developers to work on diverse hardware platforms and reduce costs for startups.
“Solving this upstream problem could unlock AI discoveries worth billions,” says Kashyap.
Kashyap believes an open framework is needed that provides a uniform development interface for AI developers regardless of the hardware they are using, similar to how Android works across different phone manufacturers. And there’s even an obvious potential partner.
“The best company actually, in my mind, positioned to work with a startup on this is Meta,” says Kashyap. It can’t be a cloud provider, because the open source AI developer ecosystem likely would be mistrustful of Google or Microsoft.
“Meta actually has a very different philosophy,” Kashyap says. “They don’t sell cloud. In fact, their whole philosophy is they want the cost of compute for themselves to come down.”
5: An AI-native app with an AI-native interface
To Tomasz Tunguz, General Partner at Theory Ventures, one of the missing pieces in AI right now is fundamentally aesthetic. If you look back to when the mobile app store came out, “everybody wrapped websites and mobile apps––and then Foursquare created the first location-native app, and everyone’s like, ‘Whoa. That’s different.’”
But we haven’t had that Foursquare moment yet.
“I don’t think we’ve really seen a native app yet,” he says. “I don’t know if we know what that looks like yet. Aside from ChatGPT, which looks like a text message from a mobile phone, we haven’t really seen an app where the UX is completely different. We’re in a phase of ‘let’s add AI here,’ let’s slap on a chatbot.”
“I think what’s missing is where there’s an innovative product manager or product designer out there, who reimagines a key piece of software and it looks literally nothing like anything we’ve seen before,” says Tunguz. “Maybe we’re missing the design paradigm.”
6: AI to make invoicing less terrible
Most B2B goods companies are still doing payments via check, says Headline partner King Goh.
“B2B business owners spend a lot of time sending manual invoices, chasing payments, reconciling the checks, and then dealing with late payments,” he says via email. “In fact, 55% of payments are late.”
What Goh suggests is a company that could be like “an AI to close the payment loop,” bringing together free check digitization services, invoicing and CRM software.
“Invoices can be automatically sent/followed-up, reconciled, etc. This allows the business owner to focus on growing their business, not just managing cash flows,” Goh says.
And the way to do that is naturally, AI.
7: AI to help mom-and-pop business better understand customers
Another B2B idea from Goh: Most B2B businesses outside professional services, especially small businesses, don’t have a CRM, so they “don’t know anything about their customers beyond hearsay/word-of-mouth, let alone grow their business using technology,” the Headline partner says via email.
“Imagine a world where a mom-and-pop manufacturing company uses AI software that knows all the safe but relevant information about their customers—payment reliability, growth potential, industry performance.” Goh muses. “Then, this information actions these insights in an automated fashion, so the mom-and-pop business can finally grow?”
8: Data is everything
If it’s got lots of data –– important data dispersed across a number of different systems –– Ken Elefant, Sorenson Capital partner, might be interested.
He’s been looking for startups in industries like “security, especially in the data center; construction, especially in commercial construction; radiology, especially in cancer-related issues; oil & gas, especially in oil exploration; and weather, especially in weather forecasting,” he said via email.
Beyond lots of data, what they have in common is this: They’re collaboration-intensive environments with “the need for a high-priced person to make sense of the disparate information,” said Elefant.
9: AI for AI
AI also needs AI. It seems like a joke but in the right application, for example, AI could help make other AI cheaper.
“The type of startup within the AI space that I think is going to be really important for the future of AI is the kind that makes AI way more efficient at the end of the day,” says Biegala, of Bison Ventures. “That means computationally efficient, that means power efficient, that means lower greenhouse gas emissions, and this could be on both the software side and the hardware side.”
Epilogue: ‘You have a lot of copycat businesses’
If there’s one thing I heard loud and clear it was this: That no one I talked to wanted to see another LLM.
“We generally look at LLMs as overfunded, and AI infrastructure is overbuilt,” said Bennett Siegel, Co-Founder and General Partner at A*. “We think the most promise is at the application layer, so that is where we spend most of our time.”
And many of the ideas above are at the application layer.
“I would say that if you’re going to go off and do something in this space, do something very different than what others are doing,” said Biegala. “It’s the type of space where you have a lot of copycat businesses, whether it’s chat bots, image generators, even reading legal documents, for example…. The last thing that you want to do is create something that’s incrementally better than a product that already exists, because you’re just going to be competed away.”
At the same time, founders out there, it’s important to know that while there’s no substitution for a great idea, differentiating completely from the rest of the marketplace is a task for Future You.
“I don’t know if everyone acknowledges that no company at the seed stage is differentiated,” said Siegel. “Now, you could have a high-caliber team that has unique advantages, relationships, and proprietary distribution––a leg up to move faster and grow more quickly. But nothing’s defensible at the seed stage. So, what we tend to look for is companies that have unique insight and have done the work to prove why they can win in this specific market.”
