How the AI gold rush is going
Over the last year, it's been a gold rush for early stage AI startups raising money. Somewhat less so for later stage AI startups who have to show the receipts. Turns out it's hard to build a differentiated AI offering that has sustainable revenue metrics.
I've watched the whole dynamic with great interest. I was an AI founder before the gold rush; maybe in some small way, Textio's early success even contributed to the gold rush. I've also spent the last several months coaching and advising numerous SaaS founders, many of whom work in AI.
Regardless of the team's specific offering or circumstances, a few themes come up over and over again in these founder conversations.
#1: Everyone is pitching job replacement.
I wrote about this a few weeks ago. Increasingly AI startups are pitching explicit job replacement in their investor decks, especially for early-career knowledge worker roles. They are also pitching benefits that amount to job replacement even if they use different words.
I already wrote about this in detail so I won't rehash it all here. The thing I want to highlight this week is how similar these pitches all sound. Most AI founders trying to raise money aren't thinking enough about this. It is difficult for investors (and potential customers) to tell them apart.
#2: AI is already a commodity. How do you build a differentiated solution?
The reason it's hard to tell AI startup pitches apart is that, in many cases, it's hard to tell the actual products apart too. Here's the situation:
- AI startups are using the same APIs to power their solutions
- AI startups are going after the same few buyers, especially in B2B
- AI startups are trying to replace the same common tasks
For instance, I'm not sure how many startups are trying to sell AI tools to write outbound sales messages right now, but I wouldn't be surprised if it was in the thousands. My Textio inbox receives ~400 prospecting messages just from people selling AI outbound sales tools every single week. And I don't even have a sales job title.
It is striking how terrible most of these messages are, which doesn't exactly inspire confidence in the solutions they're trying to sell. The messages are also all terrible in the exact same way, probably because they're all using the same APIs and prompts.
Many founders know this and are worried about it. It's hard to build a differentiated solution when everyone is using the same underlying technology and going after the same buyers and tasks.
#3: Big companies have too many advantages.
Annoying question that VCs have asked startup founders since the dawn of time: Why can't Google/Microsoft/Apple/Facebook just build this?
Historically, many founders have had valid answers to these questions. Big companies are too slow, they say. The initial market is too small for big companies to care about, they say, so we can get a great wedge before BigCo sees us coming.
But in AI, a lot of founder answers to this question are kinda bad. When it comes to AI solutions, Incumbent BigCo has two major advantages: They have workflows, and they have cash.
The cash problem is painful for startups. AI compute isn't cheap. BigCo has more financial reserves to support compute costs, sometimes without charging customers additional licensing fees for AI features. By contrast, most startups can't give AI features away for free because they just don't have the cash.
Workflows give BigCo another advantage, because people hate switching apps for related tasks. As an end user, if I can do everything in Salesforce or Workday or from my browser search bar, that's great for me. If some discrete task in my workflow can be made better with AI, I want to have the improvement right there at my fingertips where I'm already doing my work.
This is a huge issue for AI startups. To compete, they have to either build competing workflow themselves, which is expensive, non-differentiated, and time-consuming, or they have to resign themselves to being an integration feature in BigCo's app. And if the startup's AI capabilities come from APIs that BigCo can access just as easily themselves (#2 above), BigCo can choose to build the startup's feature into their platforms at any time -- and they can offer it at no additional cost.
#4: Businesses don't know how to buy AI yet.
Every B2B AI founder feels this pain. Your potential customer has convened a huge committee to make sure that the AI tools they use are secure and trustworthy. Unfortunately, the committee is generally run by a CISO or privacy counsel with no particular background in AI. The committee is doing their best with an impossible situation, but they have no idea what questions to ask to vet AI solutions.
The process for a business to evaluate a potential AI tool can take months. By the end of the process, since the committee is learning how to vet AI solutions as they go, there may be an entirely new vetting process in place. It's the worst kind of Groundhog Day.
#5: Raising the first round with an AI pitch is easy. Raising the second or third round, not so much.
You often raise your first venture round on your team and vision and market potential. You raise your subsequent rounds on your revenue, product, and market traction.
If you've recently founded an AI startup with a solid market and credible team, fundraising is still pretty good. But all of the factors above make it hard for an AI startup to get the kind of traction that investors expect for later rounds. Especially when the fundraising market is crowded with numerous other AI startups that raised monster rounds 18-24 months ago and are all grappling with the factors above.
#6 Oh, yeah. All the normal founder stuff is still hard too.
AI founders are like all other founders. The list of difficult things is long: working with investors, fundraising, hiring, prioritization, scaling, management, loneliness, pressure to grow and succeed, and lots more. Nothing new here, but it's sure a roller coaster for anyone doing it. For AI founders in particular, it's not made any easier by the factors above.
What do you think?
Thanks for reading!
Kieran
Catch up on nerd processor case studies | Subscribe | nerdprocessor.com