Just a few years ago, one of the hottest topics in enterprise software was “Robotic Process Automation (RPA).” These services, which attempted to automate many of our repetitive business processes, seemed to fall short of expectations. But the rise of generative AI may be the missing key to building such systems.
Seattle-based Tektonic is one of the newer startups in this space: The company, which combines GenAI with traditional symbolic techniques, emerged from stealth today to announce a $10 million seed round of funding led by Point72 Ventures and Madrona Ventures.
The idea here is to allow users to use natural language to interact with GenAI agents to automate workflows. One area the company is emphasizing is quotes and renewals, which often involve a series of manual tasks and are difficult to automate because each company has its own (often dynamic) processes.
Tektonic was co-founded by Nic Surpatanu, who previously held leadership roles at Tanium, UiPath and Microsoft. “Last year, generative AI emerged and we realized it was enabling some software scenarios that weren’t possible before,” says Surpatanu.[Based on] With what I've learned at UiPath and Microsoft, I know how far we can push traditional automation.”
Image credit: Tektonic
Perhaps more importantly, he says, he believes generative AI can't be treated like a magic box: “You need to combine it with symbolic methods, and if you want to get the best results, you need to combine it with more traditional software,” he says.
Surpatanu argues that generative AI can bring a previously unavailable quality to these systems: context adaptability and understanding of user intent, something that RPA often struggles with. With these older tools, any major change to the user interface breaks your scripted automations. And once you've created a set of automations, you have to commit to maintaining them.
AI also enables the extraction of semantic entities and mapping them to user intent.
“Our approach[…] “The short answer is, it's not going to be 100% flexible — we're not claiming that — but it's introducing enough flexibility to cover a much wider range of scenarios than we had before,” Surpatanu said. He noted that humans will likely remain in the loop, at least for the time being, because he doesn't believe current models are yet reliable enough to power fully autonomous agents. But he also emphasized that if tools like Tektonic can take the current state of the art from 50% automation of a process to 80%, that would be a big step forward in itself.
On the technical side, Tektonic leverages a combination of foundational and open models for entity extraction and low-level actions.
“Instead of performing manual tasks across multiple applications, salespeople should partner with AI-based agents who understand the process and get the job done, freeing them up to spend more time serving customers,” Madrona's Steve Singh, Ted Kumarat and Palak Goel said in the company's announcement today. “The emergence of generative AI models with the ability to reason across data silos within applications and orchestrate tasks allows us to rethink process automation and take it to levels never before seen.”
Tektonic is still in its early stages, and the team is currently working with several design partners to test and build the system. “In three to five years, we'll be a SaaS company, and customers will come to us and we'll connect to their system's APIs,” Surpatanu says. For now, though, getting up and running with Tektonic requires installing the system as a container on a company's virtual private cloud.