What the business is
AI platform that converts recordings or written descriptions into software workflow automations, attended or unattended
The bet
That LLMs can turn a plain video or description of a process into a working automation — replacing kludgy low-code RPA at up to 10x the speed and 1/10th the price
Starting capital
$3.75M equity round from Initialized Capital and Khosla Ventures
How it started
Lucas Ochoa and Gautam Bose were hired together out of college at Google Creative Lab and spent three and a half years shipping products including AI Test Kitchen, Teachable Machine and Pixel Buds Pro work before leaving for Y Combinator's Winter 2023 batch
What happened
Lasso rebranded to Automat; a team of three full-timers plus about a dozen contractors built each automation in up to a few days, with HubSpot and Salesforce integrations, and the new round funds a Sandbox feature and standalone document- and image-extraction APIs
What has to be true
RPA incumbents sell proprietary frameworks and licensing fees; Automat's product decision is to make the input a video or a sentence, not a flowchart.
The founders' Google LaMDA work — 50+ prototypes — is the stated origin of the conviction that LLMs should act in software, not just chat.
Pricing is the wedge: a claimed tenth of incumbent cost targets companies that could never afford classic RPA implementations.
What can be applied
When incumbents sell complexity, the wedge is the instruction format: natural language in, working automation out.
Aftermath
As of the September 2023 report, Automat had wrapped its pilot, onboarded its first paying customers, and was recruiting machine-learning engineers to balance customer implementations with R&D, in an RPA market Forrester forecast to reach about $16 billion by 2025.
FOLLOW THE EVIDENCE