Decagon raises $35 million from Accel and A16z to build AI customer support agents. Here's an exclusive look at the pitch deck it used.

"We talked to a bunch of big companies around various use cases," Zhang said. "But honestly, we just kept coming back to the customer team."

Customer support is a category ripe for change, given rising customer expectations and an increasingly competitive market. "[The best companies] understand that customer service is not a cost center, but a meaningful long-term revenue generator and retention driver," noted Kimberly Tan and Jeff Silverstein, investment partners at A16z, in a blog post about investing in Decagon

Enterprise companies either build in-house support teams or, more commonly, outsource the function. Companies like Zendesk and Freshdesk boost these support teams' efficiency by helping them create, manage, and resolve customer tickets. Many AI support tools have recently popped up to supplement or completely replace the customer support function. For example, other startups like Intercom and Forethought provide AI-based chatbot solutions to resolve customer tickets automatically.

While chatbots typically respond to customer queries with a decision tree of programmed responses, Decagon intends to go further. The company strives to provide human-like customer support with its AI agents using third-party and fine-tuned large language models. For each task, the platform tests and selects the best model. The new generation of AI support applications are "LLM models working together to figure out what needs to be done," Zhang said. "That includes taking action."

Decagon's AI agents incorporate customers' workflows and data sources, including brand guidelines and compliance requirements. This allows them to handle a wider range of support tasks, including postponing shipments, filing bug reports, and processing refund requests. "No one's even remotely tackled that in the past, just because it wasn't possible," said Zhang. "But now, with generative models, a lot of those things are possible."

Decagon also provides AI-powered analytics to spot trends and recommend operational improvements. Its vision is to help companies transform their approach to customer support from reactive to proactive.

"We've been able to quickly identify themes in our customer experiences, identify outliers, and utilize more efficient workflow processes," said Thatcher Foster, vice president of client solutions at Bilt, in a press release.

Besides Bilt, the company boasts Eventbrite, Substack, ClassPass, Vanta, and Rippling as customers. Decagon helps customers increase efficiency, reduce costs, and improve customer satisfaction, said Zhang. Customers are charged an annual fee for a set number of conversations, with additional costs for any overages. Decagon has "reached seven figures ARR very quickly," said Zhang.

"Our North Star for the company is building the most human-like solution," Zhang said. The two cofounders are on the same page about that too.

Check out the 10-slide pitch deck Decagon used to raise $35 million: