Sumble comes out of hiding with $38.5 million to bring AI-powered context to sales analytics

Ask any salesperson how much information they would love about a potential customer and you won’t ever hear the end of it. This assumption is at the heart of the crowded sales intelligence market, where today there are services available that do the whole lot from helping you discover prospects and reveal their background to writing a proposal and performing autonomous follow-up.

However, sales teams need greater than just data; they need context. Subtlestartup in San Francisco, tries to provide this context by searching the Internet through social media, job boards, company web sites, regulatory filings, etc. to obtain information about what is happening inside firms.

The brainchild of Anthony Goldbloom and Ben Hamner, founders of the data science and machine learning community KaggleSumble uses a knowledge graph powered by large language models to connect the various data points it collects. The result, Goldbloom told TechCrunch, is a comprehensive view of a company’s technology data — what tools are getting used in what departments, projects being launched or underway, the organizational chart, what technology the company might want to adopt and, most significantly, who to contact.

But given how crowded this market already is, from incumbents to countless AI sales development representatives, the query is: does the world actually need more?

Goldbloom thinks so, and says the startup’s approach seems to be working: He told TechCrunch that since the startup launched in April 2024, it has added 19 enterprise customers, including Snowflake, Figma, Wiz, Vercel and Elastic, and has tens of 1000’s of users in total. About 30% of users are paying for a Pro subscription (either themselves or their company doing so), and so far growth has been driven by word of mouth. The startup didn’t want to share its revenue details, but we understand that revenues grew by 550% year-on-year.

“What usually happens is that we go viral within the company,” Goldbloom said. “We will go from 1 to 500 MAU [monthly active users] in the company for a period of six months. Spreading typically happens inside a Slack channel, then inside one team, then inside the office, and finally inside that company.

Goldbloom said traction, customer quality and high customer retention played a big role in attracting investor attention. On Wednesday, the startup emerged from obscurity with $38.5 million in funding — Coatue led an $8.5 million seed round, while Canaan Partners led a $30 million Series A round. AIX Ventures, Square Peg, Bloomberg Beta, Zetta and angel investors including Salesforce CEO Marc Benioff and former GitHub CEO Nat Friedman also invested.

It’s price noting that Sumble’s co-founders have attracted investors they know well. Rich Boyle, now a general partner at Canaan, was a board observer at Kaggle, while Bloomberg Beta and Zetta also sat on Kaggle’s cap table. Goldbloom also co-founded AIX Ventures and is an investment partner, although he told TechCrunch that he “walked out of the room” when the company was considering investing in Sumble.

Still, Sumble faces a lot of competition. Competitors include Apollo.io, Slintel, SalesLoft, Cognism, Reply.io, ZoomInfo, HubSpot and Outreach, which give either more focused point solutions or comprehensive sets of IT sales tools. And because Sumble now uses publicly available data, there’s nothing stopping others from doing what they’re doing now.

However, Goldbloom believes that Sumble’s moat is deeper than it appears at first glance, thanks to the structure of its knowledge graph, which covers roughly 2.6 million firms worldwide.

“The way we think about it is that the more data we add to the knowledge graph, the richer the corpus will be. We see the richness of the knowledge graph as a huge source of defense opportunities,” he said.

Sumble is also counting on the continued adoption of large language models, which is able to help it proceed to scale because it expects people to use AI along with its services. “The way we structure our data makes the knowledge graph is and always will be easily searchable by large language models […] “The idea is you possibly can ask ChatGPT about Apple’s text stack, or you possibly can ask ChatGPT about Apple’s technology stack based on our data,” Goldbloom said.

“We believe that AI will greatly change the data provider landscape in such a way that having a knowledge graph structure as a way to provide context to a large language model will be a key part of the LLM ecosystem,” he added.

The service is currently offered as a web application and via API. There is also a paid plan available that provides more features, resembling integration with workflows and CRM systems, in addition to notifications when a lead development may be of interest.