Most automation software starts with a blank screen and a blinking cursor, which is a problem. The people who know the work best are not the people who know how to script it. Automat, a startup founded by two Google Creative Lab alumni, starts instead with a screen recording. You show it the work, and it builds the bot. It is a simple idea with a complicated goal: to replace the brittle, scripted automation of the past with something that can understand context, recover from errors, and, in theory, improve [TechCrunch, 2023].
This approach has now secured the company a $15.5 million Series A round led by Felicis Ventures, with participation from Khosla Ventures, Initialized Capital, K5 Global, and Input Capital [Axios, 2025]. The funding follows an earlier $500,000 seed from Y Combinator in 2022 [Perplexity Sonar Pro Brief]. For a company in the crowded field of AI automation, the round is a vote of confidence in a specific wedge: using video as the primary input to tame enterprise complexity.
The video wedge
The core of Automat's bet is that demonstrating a workflow is more intuitive than describing it. The company's platform, which was originally called Lasso, allows a user to automate a task by sending a screen recording of themselves performing it or by writing a plain text description [TechCrunch, 2023]. The system then generates an automated workflow, or 'agent,' designed to replicate the task.
This is positioned as a direct challenge to traditional robotic process automation (RPA). Legacy RPA tools are famously fragile; they follow rigid, pre-programmed paths and break when a button moves or a form field changes. Automat's agents are meant to be more adaptive, using AI to interpret on-screen context and navigate unexpected changes [Perplexity Sonar Pro Brief]. The company now frames its offering around two product lines: an 'AI employee' named Ace for more autonomous tasks, and a managed service where its team builds and maintains automations for enterprises [Automat, 2026].
Founders from the prototype shop
The founders, Lucas Ochoa and Gautam Bose, are a study in product-focused partnership. They met as students at Carnegie Mellon University and were hired together out of college to join Google Creative Lab [TechCrunch, 2023]. For three and a half years, they worked on launching experimental products, including the AI Test Kitchen, Teachable Machine, and the Google Pixel Buds Pro [TechCrunch, 2023]. That background in rapid prototyping and user-centric design is evident in Automat's video-first approach. Ochoa serves as CEO, with Bose as Chief Strategy Officer [LinkedIn, 2026].
Their experience shipping consumer-facing AI tools at Google gives them a different pedigree than the typical enterprise software founder. They are not coming from a background in selling six-figure RPA contracts to Fortune 500 procurement offices. Instead, they are applying a builder's mentality to a staid enterprise problem, betting that ease of use can be a moat.
The funding trajectory and competitive silence
Automat's funding history shows a significant step-up from its Y Combinator seed to its recent Series A, though the exact valuation is not public. The investor list is notably stacked with firms known for backing technical founders and category-defining bets.
2022 Seed | 0.5 | M USD
2025 Series A | 15.5 | M USD
What is absent from the public record is almost as telling as what is present. The company has not disclosed named enterprise customers, specific deployment counts, or detailed revenue metrics. In a market where traction is often measured by logos on a slide, this silence is a strategic choice or a reflection of early stage. The competitive landscape is also vaguely defined in sources, with no direct rivals named. This suggests Automat is either carving a unique niche or has yet to collide head-on with the established giants in the automation space.
Where the wheels could come off
The ambition is clear, but the path is lined with execution risks that go beyond software development. Automat is making several high-stakes bets simultaneously.
- The complexity ceiling. Automating a five-step process from a video is one thing. Automating a 50-step, cross-application workflow with conditional logic and exception handling is another. The company's promise of agents that 'understand context and recover from errors' must hold under the weight of real enterprise spaghetti [Perplexity Sonar Pro Brief].
- The go-to-market climb. The team's product strength is an asset, but enterprise sales require a different muscle. Winning and expanding within large organizations, navigating IT security reviews, and displacing entrenched RPA vendors like UiPath or Automation Anywhere is a brutal, relationship-driven slog.
- The economic model. The company offers both a self-service tool and a managed service [Automat, 2026]. The managed service can drive high value but does not scale like software. The self-service product must prove it can capture enough value to justify its cost against a sea of other automation and AI co-pilot tools.
The company's most plausible answer to these risks is its focus on the video wedge. By drastically lowering the barrier to creation, they hope to ignite automation projects that would never have been greenlit under the old, expensive, and technical model. Success would be defined not by beating a legacy vendor in a head-to-head feature war, but by growing a new market of automators entirely.
The next twelve months
The $15.5 million is likely earmarked for scaling the team and proving the model with early lighthouse customers. The key milestones to watch will be less about raw customer count and more about depth of use. Can Automat point to a Fortune 500 company that has moved a meaningful, business-critical process onto its platform? Can it show that an automation built via video is not just easier to create, but also more robust and cheaper to maintain over a six-month period than a traditional RPA script?
The unit economics of automation are brutally simple. The value of a bot is the fully-loaded cost of the human labor it replaces, minus the cost to build and maintain the bot itself. If a finance clerk earning $65,000 a year spends a quarter of their time on a report that Automat can handle, the annual value is about $16,250. If Automat's service costs $5,000 a year to build and run, the net saving is $11,250. The company wins when its share of that saving exceeds its own cost of delivery. The incumbent it must beat isn't just another software vendor, it's the internal spreadsheet macro, the outsourced offshore team, and the stubborn belief that 'the way we've always done it' is cheaper than change.
Sources
- [Perplexity Sonar Pro Brief] Automat company profile
- [TechCrunch, 2023-09-06] Y Combinator-backed Automat turns videos into workflow automations | https://techcrunch.com/2023/09/06/y-combinator-backed-automat-turns-videos-into-workflow-automations/
- [TechCrunch, 2023-04-04] These Y Combinator-backed startups are trying to build 'ChatGPT for X' | https://techcrunch.com/2023/04/04/these-y-combinator-startups-are-trying-to-build-chatgpt-for-x/
- [Axios, 2025-11-19] Automat raises $15.5M led by Felicis to help automate workflows | https://www.axios.com/pro/enterprise-software-deals/2025/11/19/automat-15-million-enterprise-workflows
- [LinkedIn, 2026] Lucas Ochoa - CEO @ Automat | ex: Google & Microsoft
- [LinkedIn, 2026] Gautam Bose - CSO @ Automat | Ex-Google Creative Lab
- [Automat, 2026] Automat | AI Teammates and Managed Automations | https://runautomat.com/
- [Y Combinator, 2026] Automat: Agentic automation that understands context, recovers, and improves. | https://www.ycombinator.com/companies/automat