Clay hits 10x growth by turning GPT-4 into an army of AI SDRs

Most sales teams are drowning in tools that don’t talk to each other. One platform for prospecting, another for enrichment, a shaky spreadsheet for everything else. Clay’s bet was simple: what if you could collapse all of that into a single AI agent that actually works?

They built Claygent on GPT-4, and the results are frankly hard to ignore. The company has logged 10x year-over-year revenue growth for two years straight, then accelerated to 2.5x growth in just the first five months of 2024. Over 100,000 users are on the platform now, including teams at Intercom, Verkada, and Notion. Kareem Amin, Clay’s co-founder and CEO, didn’t mince words about the old way of doing things: “All tools before LLMs tried to automate what SDRs were doing in terms of research but they would get stuck.” His point was that legacy providers capped out at 100 data points per company, and custom scrapers were a maintenance nightmare.

What makes Claygent different isn’t just the model — it’s the engineering around it. Rather than blindly feeding entire websites into GPT-4 and burning tokens, the agent first asks the model where the target information is most likely to live. A SOC-2 report? Probably the footer. Then it uses a binary search pattern to progressively narrow the search space, slicing and checking sections until it finds the right data. For simpler tasks like their AI Formula Generator, Clay defaults to cheaper, smaller models, reserving GPT-4 Turbo for the heavy cognitive lifting. They also cross-verify scraped data against multiple providers, which is the kind of belt-and-suspenders approach you need when sales teams are betting pipeline on these insights.

Adoption numbers tell the real story. Thirty percent of Clay’s customers use Claygent daily, generating 500,000 research and outreach tasks every 24 hours. An ecosystem of “Claygencies” — micro-agencies often founded by former SDRs — has sprung up to offer data enrichment, lifecycle campaigns, and SEO services entirely within Clay. Teams of one or two people are now doing the work that used to require a full agency. Clay’s roadmap points toward fine-tuned agents for verticals like healthcare and fitness, plus proactive triggers that could detect when a prospect company just hired ten engineers. If that sounds like a fundamental shift in who needs to be in the room for GTM execution, that’s because it probably is.