AI “agents” are generative AI models capable of performing tasks autonomously, such as copying information from an email and pasting it into a spreadsheet. While they are often praised as productivity enhancers, their tendency to make mistakes suggests this acclaim might be premature. Nevertheless, some founders, analysts, and investors believe that agents represent the next frontier in generative AI.
Bella Liu and William Lu are among these founders. Their company, Orby AI, is developing a generative AI platform designed to automate various business workflows, including data entry, document processing, and form validation.
Many startups offer tools to automate repetitive back-office tasks, such as Parabola, Tines, Induced AI (backed by Sam Altman), and Tektonic AI. Established companies like Automation Anywhere and UiPath are also incorporating AI to stay competitive in the generative AI space.
Liu and Lu assert that Orby’s technology is unique in its ability to learn and act on workflows in real time, understanding patterns and relationships within an enterprise’s unstructured data.
“Orby’s platform observes how workers perform their tasks to automatically create automations for complex activities that require some level of reasoning and understanding,” explained Liu, Orby’s CEO. “An AI agent installed on a worker’s computer effectively watches, learns, and generates automations, adapting as it learns more.”
Orby, which emerged from stealth in 2023, aims to develop AI that can comprehend and abstract low-level decisions made by workers, allowing them to focus on more complex tasks.
Liu previously led AI and automation initiatives at IBM, including product planning and AI-related mergers and acquisitions, and served as UiPath’s director of AI product management. Lu, a former Nvidia systems engineer, joined Google Cloud as an engineering lead, where he helped design generative AI document and database extraction technologies.
Orby’s unique advantage lies in its cloud-based generative AI model, which is specifically fine-tuned to handle customer tasks like validating expense reports. This model partially utilizes symbolic AI, a type of AI that applies rules, such as mathematical theorems, to deduce solutions to problems.
Symbolic AI on its own can be rigid and slow, especially with large and complex data sets, as it requires well-defined knowledge and context to function effectively. However, recent research indicates that it can be scalable when combined with traditional AI model architectures.
“For the past two years, we’ve been developing this AI model and have conducted successful trials,” Liu said. “There are few pure-play generative AI companies tackling the enterprise market with a comprehensive solution. We are one of them.”
Liu explains that Orby’s model can intelligently adapt to workflow changes, such as updates to an app’s UI, by analyzing API interactions and a worker’s browser usage. While monitoring an employee’s every move might sound like a privacy concern, Liu assures that Orby does not store most customer data. It only uses specific telemetry data to enhance its model, with encryption applied both in transit and at rest.
“Humans are kept entirely in the feedback loop,” she added.
Orby recently secured $30 million in a Series A funding round co-led by New Enterprise Associates, Wndrco, and Wing, reportedly valuing the company at $120 million post-money. The company is competing in a challenging sector, with upcoming agentic AI from major players like OpenAI and Anthropic affecting the prospects of both incumbents and smaller companies.
Adept, a startup focused on AI agents for enterprise applications, is reportedly nearing an acquihire deal with Microsoft before launching a single product. Amazon and Google have released AI agent tools with little fanfare. Meanwhile, UiPath, despite increasing its generative AI efforts over the past year, saw a decline in sales in its most recent fiscal quarter.
Liu believes Orby can succeed by adopting a systematic go-to-market strategy. The company is already generating revenue from about a dozen customers and plans to use its $35 million fund to expand its Mountain View-based team of roughly 30 people.
“The funds are being used to scale our go-to-market, customer support, product, and technical organizations,” she said. “The enterprise market has a strong demand for generative AI solutions that significantly improve business performance. They are just trying to determine the best applications for the technology in the near term before scaling it across their operations.”