We believe in the future Thread AI is building—and we’re proud to be part of their Series A.
Over the past year, AI agents have demonstrated their potential to automate domain-specific workflows in areas like medicine, law, and software development. However, most businesses also rely on bespoke processes that require customization beyond what is available in out-of-the-box solutions. In the past, businesses looking to implement their own automations typically looked to RPA, but today, they want the more intelligent, flexible automations that AI agents promise, and these capabilities quickly exceed the limits of traditional tools.
In practice, AI-powered workflows can vary in scope, but they generally involve a step where a model is called to reason over data, generate additional information, or dynamically plan and execute based on inputs. In other words, they involve the AI assuming responsibility for some sort of “cognitive” load. Some use cases are as straightforward as aggregating data from siloed SaaS tools, calling an LLM to reason over the data, and updating a database given the output. Others quickly get more complicated and may even involve physical systems—for example, ingesting hardware sensor logs, detecting anomalies, developing a triage plan, and handing a support ticket off to a human.
While at first, these implementations may sound like a close derivative of RPA tools or even the recently popular agent builders, in reality, they require a much more sophisticated backend to remain reliable and secure. For example, if a workflow fails due to a model making an erroneous decision, that not only needs to be flagged to a human, but the system should also be fault-tolerant and capable of re-executing the task or modifying itself until it succeeds. This type of orchestration is known as durable execution, and it is a problem space that is already notorious for its complexity and operational overhead. Once you introduce added layers like human-in-the-loop where agents must pause execution and maintain state and context for indeterminate periods as they wait for input, the challenges multiply. Most existing tools weren’t designed with these needs in mind, and that gap has become a significant barrier for organizations.
Thankfully, this is where Thread comes in. Rather than hiring high-end developers to build and maintain durable execution systems or trying to mangle your workflow into the brittle RPA tools, Thread’s workflow orchestration platform, Lemma, is purpose-made for building these agentic workflows in an accessible, robust way.
Rather than manually chaining together disparate enterprise systems and data sources, Lemma unifies the control plane, enabling AI integration without the heavy lifting of infrastructure management. Customers are using Thread to automate tasks like research, proposal generation, and incident reporting where messy, often multi-modal data is gathered and reasoned over from several sources.
Thread’s co-founders, Angela and Maya, bring extensive experience from their time leading AI product development at Palantir. Angela, formerly Head of AI Product, and Maya, former Head of AI/ML Engineering, played pivotal roles in evangelizing and developing Palantir’s AI strategy and making it a core part of the company’s platform. They have an indisposable level of technical experience and a deep understanding of the implementation challenges that these large-scale enterprises face, and they are now bringing that to their work at Thread.
We are entering a world where AI’s impact on the way businesses operate is undeniable. Teams that free up their time via automation will experience leaps in productivity, and those that don’t will struggle to keep pace. However, none of this can happen without platforms like Lemma. We are thrilled to partner with Angela, Maya, and the entire Thread AI team as they continue to define how enterprises will leverage AI to drive efficiency and innovation.
