The most expensive part of a SaaS sale happens after the contract is signed. Customer success teams spend hours clicking through their own software, building walkthroughs, writing documentation, and chasing down bugs. SAMMY Labs, a London-based startup that came out of Y Combinator's W25 batch, is betting an AI agent can do that work by exploring an application like a real user [Perplexity Sonar, 2025]. The company's engine simulates clicks, maps flows, and generates personalized content, aiming to automate the post-sales grind for software companies [Crunchbase].
The bet on screen-aware agents
Most automation tools work on the back end, via APIs or databases. SAMMY's differentiation is its focus on the user interface itself. Its AI is described as "screen-aware," meaning it can navigate an app's front end, identify UI elements, and perform actions to learn the product [SAMMY Labs Docs]. This allows it to generate up-to-date walkthroughs and documentation that reflect the current state of the software, a task that otherwise requires constant manual updates from human teams. The core promise is continuous UI testing and content generation without manual effort [Perplexity Sonar, 2025].
Early traction and the YC stamp
Founded in late 2024 by Joe Savidge, Shav Vimalendiran, and Joshua Carey, SAMMY Labs moved quickly into Y Combinator [Preqin, March 2025]. That accelerator pedigree provided initial momentum and was followed by an undisclosed seed round led by Pioneer Fund earlier this year [Preqin, March 2025]. The company reports it is already running across more than five verticals, with customers ranging from seed-stage startups to firms valued at over $10 billion [Perplexity Sonar, 2025].
| Founder | Role |
|---|---|
| Joe Savidge | Co-Founder |
| Shav Vimalendiran | Co-Founder & CTO |
| Joshua Carey | Co-Founder & Founding Engineer |
Where the execution gets hard
For all its promise, SAMMY's bet rests on solving several non-trivial technical and commercial challenges. The accuracy and reliability of a UI-exploring AI in complex, permissioned enterprise environments is unproven at scale. Furthermore, the value proposition must be compelling enough to displace existing workflows built on tools like Pendo, Zendesk, or dedicated QA testing software.
- Technical fidelity. An agent that misclicks or misinterprets a dynamic UI could generate incorrect guides or miss critical issues.
- Integration depth. To be truly useful, the agent likely needs to pull context from sources like Slack threads and support tickets [Perplexity Sonar, 2025].
- Defensible differentiation. SAMMY will need to prove its screen-aware approach creates a tangible efficiency gap over competitors that might use simpler screenshot analysis or API-based methods.
The next twelve months
Over the coming year, watch for SAMMY Labs to move from a promising prototype to a hardened enterprise product. Key milestones will include announcing its first publicly referenceable customers and detailing its pricing and packaging. The company will also need to demonstrate that its AI can handle the long-tail of customizations and edge cases present in its target $10B+ customer environments.