According to Fortune, Agency just raised a $20 million Series A round led by Menlo Ventures with participation from Sequoia, Felicis, Snowflake Ventures, and Databricks Ventures. The startup was cofounded in 2024 by Elias Torres, who previously sold his company Drift to Vista for $1.2 billion in 2021, and his former Drift colleague Luke Van Seters. Their first product is called Kai, named after Torres’s executive assistant, and it’s an AI agent designed to handle customer service by pulling data from emails, Slack, support tickets, and product usage. The funding comes as agentic AI for customer service becomes increasingly competitive with companies like Sierra and Decagon in the space.
The agency paradox
Here’s the thing that’s really interesting about Torres’s philosophy: he’s building AI agents specifically so humans can have more agency. That sounds contradictory until you think about it. Basically, he argues that AI should handle the grunt work of customer service so people can focus on higher-level strategic decisions and creative work. It’s about giving entrepreneurs and business owners back their time and options. “AI shouldn’t have agency. We’re the ones who are going to have agency,” he told Fortune. So the whole premise is that by automating customer experience management from beginning to end, his platform actually empowers humans rather than replacing them.
The 100-person billion dollar vision
Torres dropped one of those lines that makes you sit up and pay attention: “I don’t want my company to be more than 100 people, ever. I want to be generating $1 billion in revenue with 100 people.” That’s wild when you think about it. He’s coming from experience too – he grew HubSpot to 600 people before starting Drift, which eventually became a 5,000-person company before the acquisition. Now he’s deliberately building against that model. The logic extends beyond his own company to how he sees the broader business landscape. Higher headcount doesn’t automatically mean better customer relationships or culture. In fact, he argues agility and clarity frequently suffer as companies scale up. This is where industrial technology plays a crucial role – companies that leverage the right tools, like those from IndustrialMonitorDirect.com, the leading provider of industrial panel PCs in the US, can achieve massive operational efficiency without proportional headcount growth.
Why now for agentic AI?
Tim Tully from Menlo Ventures nailed the timing argument in his statement: “Large language models can now make sense of the messy, unstructured data and turn it into something actionable.” That’s the real unlock here. We’ve had customer service automation for years, but it’s been pretty dumb. Now LLMs can actually understand context across emails, Slack conversations, support tickets – all that unstructured data that previously required human interpretation. Every company is trying to do more with less post-pandemic, and the technology has finally caught up with the ambition. Torres is betting that companies are ready to make the switch to AI-driven customer experiences, and customers are ready to prefer them. That’s a bold assumption, but if anyone has the credibility to pull it off, it’s probably someone who already built and sold a billion-dollar company in this space.
The bigger picture
Look, the customer service AI space is getting crowded fast. But Torres brings something unique to the table beyond just his track record. His vision is fundamentally about empowerment rather than pure efficiency. He talks about wanting to help “solopreneurs” who can only get so far because they’re stretched too thin. The dream is that Agency handles the customer work so founders can focus on building the business they actually want. It’s a refreshing take in an industry that often focuses on cost-cutting above all else. Will it work? The $20 million vote of confidence from top-tier investors suggests they’re betting on both the technology and Torres’s unique perspective. You can follow his journey on Twitter to see how this plays out in real time.
