Lucas Ochoa and Gautam Bose were hired together out of college at Google Creative Lab, Google's tech and culture incubator, and spent three and a half years launching products like the AI Test Kitchen, Teachable Machine and Google Pixel Buds Pro. In early 2020, their team was tasked with finding consumer applications of Google's LaMDA large language model, and they built more than 50 prototypes integrating LaMDA across Google’s apps and services.
While working with LaMDA, Ochoa and Bose say that they came to realize that enterprise automation relies too heavily on outdated tools burdened with licensing fees, and that many of the tools meant for "citizen developers" turn out to be more complex than necessary.
"The industry's main challenge is integrating the advancements brought by generative AI, essentially finding the most user-friendly automation methods using AI," Ochoa and Bose told TechCrunch in an email interview. "We have conviction that solving the action side of software automation will be critical to giving future, more sophisticated AI models proper agency to do valuable work."
Drawing inspiration from robotics techniques that tap large language models like LaMDA to translate instructions into actions, Ochoa and Bose built an AI system -- Lasso -- designed to translate a video or description into a software automation workflow. Ochoa and Bose eventually left Google to work on Lasso full time, joining Y Combinator's Winter 2023 batch.
Lasso, which recently rebranded to Automat, can break down a recording of a workflow into step-by-step instructions. Then it can apply AI to interpret the meaning of those instructions -- performing actions on a user's behalf.
Ochoa and Bose claim that Automat can automate essentially any workflow on a PC, from processing insurance claims to streamlining the submission of construction permits and submitting trade licenses. Leveraging AI and integrations with a range of third-party apps, including HubSpot and Salesforce, the platform can parse documents, execute inventory management and even respond to text messages and emails.
Automat comes in two flavors -- unattended and attended. The unattended version runs in the cloud and doesn't require human oversight, while the attended version acts like a "copilot" alongside a user to assist with repetitive tasks.
"Put simply, Automat takes care of repetitive, boring and dull work that prevent businesses from scaling," Ochoa and Bose said. "We collaborate with executives and engineers that are looking to build efficiencies into their operations or use automation to enable wholly new product offerings."
Now, Automat isn't magic. Contrary to what this reporter initially assumed, the platform doesn't translate videos into automations instantly -- or even automatically. Customers have to submit a video or description of the process they want to automate to the Automat team, which takes up to a few days to build the automation.
Automat's platform to automate repetitive tasks. Image Credits: Automat
Automat's also not the first to market with a video-to-workflow conversion tool. A number of incumbent robotic process automation (RPA) vendors offer comparable solutions, including MuleSoft, Microsoft (in Power Automate) and ServiceNow.
But be that as it may, Ochoa and Bose claim that Automat is up to 10x faster and 1/10th the price of the workflow automation competition.
"The dominant players in the space are companies like UI path, Intelligent Automation and Blue Prism. Our approach is different; instead of kludgy low-code tools or proprietary programming frameworks, we offer powerful automations built with simple instructions," they said. "For companies that don't have the means to scale, we enable them to affordably eliminate time-consuming tasks so that they can focus their energy on what matters."
Automat recently wrapped up a pilot program and is working with its first batch of paying customers, which Ochoa and Bose claim range from small- and medium-sized business and "large-scale enterprise" corporations.
To grow the platform, Automat recently raised $3.75 million from Initialized Capital and Khosla Ventures in an equity financing round. The proceeds will be used to support the launch of Automat's new Sandbox feature, Ochoa and Bose say, which will let users manage workflows and access new standalone document and image data extraction APIs. A portion of the funding will also be put toward expanding Automat's workforce from three full-time employees and around a dozen contractors.
"We're using this funding round to grow our engineering team and are actively recruiting for folks with machine learning experience. We decided to raise based on our early traction and demand from our customers," Ochoa and Bose said. "With this boost, we’re building a lean technical team that will balance customer implementations and an ambitious R&D roadmap."
It's a good time to raise. The market for RPA -- tech to automate repetitive, mundane enterprise tasks -- appears to be stabilizing after a long period of stagnation and decline. According to research by Forrester, the segment will grow to around $16 billion by 2025.
"Automat is fortunately the kind of tool that’s necessary in all weather: Through the market’s inevitable ebbs, companies rely on automations to save operational cost and increase efficiency. On the other hand, during periods of growth, flow automation lets companies grow unencumbered by expensive scaling costs."