When you really probe venture capitalists about investing in AI startups, they’ll tell you that businesses are experimenting wildly but are very slow to add AI solutions into their ongoing business processes. 

But there are some exceptions. And one of them appears to be an area known as AI sales development representatives, or AI SDRs. These use large language models (LLMs) and voice technology to craft personalized outreach emails and place automated calls to potential customers. 

“In some markets, we’re seeing five to 10 companies all have success in a pretty short period of time,” Shardul Shah, a partner at Index Ventures, said of the AI SDR boom.

While it’s certainly not uncommon for multiple startups to target the same problem, it’s rare to see all of them experience rapid growth. But that’s apparently the case for startups that automate content creation for sales teams, investors say.

“When one studies any of [these startups] individually, it’s like ‘wow, that’s stunning product market fit,’” Shah said. “When all 10 of them have stunning product market fit, it’s hard to answer ‘How is that going to play out?’”

Index has yet to invest in any of these companies, many of which are less than a year old. Even though the whole category is on fire and customers are using them, it’s still too early to know whether their growth will continue over the long term or whether they will be discarded like so many other AI pilot projects once the wow factor fades, because they don’t prove to be more effective than human outreach.

Small businesses love AI sales LLMs

Arjun Pillai, founder of Docket, a startup that builds AI sales engineers, is convinced that AI SDR adoption is high because small and medium-sized businesses can easily experiment with these tools. Before Docket, Pillai was the chief data officer at sales lead generation platform ZoomInfo.

“Over the last two years, the reply rate on cold emails fell at least 50%,” Pillai said. “Now that there are a bunch of companies that claim they can improve this rate, everyone is willing to try their service.”

The best-known AI SDR startups include Regie.ai, AiSDR, Artisan, and 11x.ai, but ZoomInfo, an incumbent, also released a copilot that competes with these and other virtual sales agent startups.

While these companies are experiencing rapid revenue growth, it’s unclear if they’re actually helping businesses sell more effectively.

According to Tomasz Tunguz, founder of Theory Ventures, a chief revenue officer from a publicly traded company disclosed to him that while an AI SDR helped generate a substantial volume of leads over a nine-month period, it did not lead to actual sales.

“So it’s not to say that AI won’t work. It’s to say many of us [still] do not know how to use AI,” Tunguz said onstage at a SaaStr conference in September.

Will the incumbents squash them?

Chris Farmer, partner and CEO at venture firm SignalFire, said he believes that AI applied to sales and marketing is a large opportunity, but without access to differentiated data, AI SDR startups risk being overtaken by incumbents like Salesforce, HubSpot, and ZoomInfo. Those companies’ main products are the keepers of their customers’ data. So if they offered bots that let their customers tap into their own data, such bots could be more effective. 

Another venture capitalist who looked at this market but hasn’t yet invested said her firm looked at several AI SDR startups and that they all had $1 million in ARR within less than a year. The startups’ impressive growth was attractive, she said, but like Farmer, she was concerned their solutions could eventually be offered as a free feature by established competitors. 

Jasper, a copywriting startup that was last valued at $1.5 billion but ran into speed bumps and had to lay off 30% of its staff after ChatGPT was introduced, serves as a cautionary tale for some investors. 

Investors are not surprised by the rapid adoption of AI SDRs; they are just doubting that adoption is sticky.

Updated: This story was originally published on August 22 and was updated December 26 with comments from Tomasz Tunguz.