About a year ago, TechCrunch published an investigation into 11x — at the time, one of the most hyped AI SDR startups on the market. The company had raised $76M from a16z and Benchmark on the promise of replacing human salespeople with an AI agent. The investigation uncovered fabricated customer logos and inflated ARR numbers.
I dug into it at the time. My takeaway wasn’t just about the scandal: I thought the founder didn’t seem to understand how cold sales works in real world, and was building a product for investors rather than for customers.
For context, I’m Rinat — I run getsally.io, a B2B outbound agency working with 25+ US and EU teams. We build and operate cold sales systems daily, so I follow the AI SDR space closely. A year later, I wanted to check in on the category and see what actually changed.
The short answer: the correction is underway.
The state of AI SDR market
Over the past two years, more than $400M has gone into AI SDR startups. There are now over a hundred companies in the space, and roughly three-quarters of them promise fully autonomous operation with no human involvement.
Most of the published research on the category comes from the companies selling the software. More independent sources paint a different picture: average annual churn sits at 50–70%.
What happened to the flagships
/ 11x went through a major shift. The founder stepped down as CEO. The company is rebranding from “AI SDR” to “GTM platform” and now screens clients aggressively — they require a mature GTM organization with a dedicated process owner, and won’t sell to teams where the CEO or CRO would be running the process themselves.
The new CEO — formerly the CTO — put it bluntly in a recent interview: “I don’t think AI SDRs work in the current form. And you’re hearing that from the CEO of the company that invented the word.”
/ Artisan raised $40M and made headlines last year with “Stop Hiring Humans” billboards across San Francisco. They started with the ambition of building a marketplace of AI agents for every business function, then pivoted hard into a single vertical: AI BDR. The CEO admitted to TechCrunch that early versions of the product suffered from hallucinations and that the company had been selling to the wrong customers.
I looked at their published case studies. SumUp, their largest listed client, sent over 400,000 emails through Artisan and reports 8–15 positive replies per week. Even on a conservative estimate, that’s an order of magnitude below the typical 1–2% reply rate you’d expect from well-built manual outreach on a clean list — volume compensating for what accuracy can’t deliver. The claimed cost per lead is $52, but there’s no definition of what counts as a “lead”: a reply, a booked meeting, or a qualified opportunity. And despite the category positioning itself for enterprise SaaS, there isn’t a single large enterprise SaaS client in Artisan’s public portfolio.
/ Salesforge started as an AI copilot and later released Agent Frank — a fully autonomous email agent that finds leads, writes copy, and sends messages on its own. But even favorable reviews note that a human is still needed for handling replies and non-standard situations.
The pattern
The AI SDR category accelerated on early AI hype. ChatGPT had just gone mainstream, and the market didn’t yet have a clear sense of what AI tools could and couldn’t do. Promises of cheap AI employees landed perfectly in a FOMO-driven buying environment.
A year and a half later, it’s becoming clear that turnkey AI SDR doesn’t work at the product level — not yet. Model quality and agent architecture aren’t there. The context window required for B2B outbound is massive, and the landscape shifts constantly.
B2B lead generation is a chain where every link directly affects the outcome: research, contact sourcing, segmentation, messaging, timing, post-reply follow-up. Current models can’t execute each of these steps well inside a single horizontal product.
What works instead
The practical path right now is one of two things: build your own stack of AI agents dedicated to each stage of the outbound process, or find vertical tools that solve one specific problem well. In both cases, the final calls stay with a human.
I’m confident this is a matter of time — AI agents will get there. But “eventually” and “now” are different things, and the gap between the two is where most of that $400M went.
😉 At getsally.io, when we're hiring in-house, we score every candidate on a separate column: "4o-mini," "Opus 4.6," and so on. The rule is simple — we only hire people who rate above the current flagship model.
P.S. If you’re running cold sales right now and struggling to cut through the noise, grab 30 min on my calendar. Let’s see if we can help with that.
Explore more case studies from SaaS and enterprise teams and see how structured outbound actually scales.
