Agentic AI continues to bring in the big investment bucks.
On Tuesday, data platform Hightouch announced it closed $80 million Series C funding at a $1.2 billion valuation. The round was led by Sapphire Ventures with participation from NVC and existing investors Amplify Ventures, ICONIQ Growth and Bain Capital.
Hightouch’s business previously revolved around its “Composable CDP,” a type of customer data platform that plugs directly into a marketer’s existing data infrastructure. (By contrast, traditional CDPs are intended to serve as the primary data storage system.)
But this latest funding round revolved more around Hightouch’s newest offering, an “AI Decisioning” product that uses an artificial intelligence agent to send personalized marketing materials to consumers and learn from the outcomes of that marketing.
Must love dogs
Marketers, said Hightouch Co-Founder and Co-CEO Kashish Gupta, are “scared” of generating creative from scratch these days. They’re under immense pressure to make sure their brand is well-represented, particularly when the wrong word or image might alienate their target audience.
AI Decisioning, then, is intended to remove the guesswork with reinforced learning – an advanced form of trial and error that’s typically used by recommendation algorithms – to deliver 1:1 personalized messages at scale.
The platform delivers these messages using a set of explicit rules and guardrails inputted by the advertiser, combined with the brand’s available data on customer behavior and prior ad campaigns’ success. Then, it creates and sends out messages, such as in-app promotions, text messages, emails and ads designed for walled gardens.
One example Gupta gave revolved around one of Hightouch’s largest brand clients, PetSmart. The AI Decisioning agent might source images of dogs from a preexisting database of brand content, and then home in on a specific dog breed if a user begins to respond positively to it.
“Responding positively,” in this case, means the user is performing actions the advertiser has classified as positive within the platform. The AI agent will prioritize guiding the user toward the outcomes ranked as “best” (like “purchase”) and away from the “worst” ones (like “unsubscribe”).
The platform can also reportedly operate without any human supervision, although there are manual fail-safes just in case. But for Gupta, “the real innovation is in the intelligence and decisioning layers, rather than the campaign execution layers.”
It comes recommended
AI Decisioning’s optimization and pattern recognition abilities are the result of a proprietary in-house AI model, based on the “early days” of Amazon’s product recommendation system, Gupta said. (Some of Hightouch’s engineers originally worked on that system, too, in fact.)
While many AI systems operate as a black box, Hightouch designed its platform to share with marketers what’s working. They can see a wide range of variables that contribute to success, such as user location or the last page they clicked on. This transparency can help marketers come up with better instructions to feed to the AI or generate new ideas for campaigns to run on.
For example, if a significant number of users from Texas start responding to marketing content with outdoor imagery in it, a savvy marketer might decide to start unique campaigns for that part of the country.
“If you didn’t have that learning, you might just keep optimizing,” said Gupta. “And maybe the AI is getting smarter and smarter, but you as a marketer are not learning from that.”
The platform certainly has Hightouch’s investment partners excited, at least. This Series C funding round, which closed in less than three weeks, was borne primarily from investor interest, said Gupta. Sapphire Ventures Partner Rajeev Dham, in particular, has been following the company for years and will now be joining the Hightouch board.
With this new $80 million, Hightouch plans to double its engineering headcount and beef up its sales and customer service teams for better customer acquisition. But in the long term, Gupta hopes to find further use cases that can carry on beyond AI’s current buzziness.
“We would love to call it agentic marketing,” he said. “That is what we think we are building and will continue to build for many years.”
