Sonder Labs
Sonder is an applied AI lab that builds digital twins of individual workflows by using vision models to watch a user's screen and learn their patterns, preferences, and work habits. The platform turns recurring work into reusable, shareable skills that can be deployed promptlessly, with the goal of scaling from individual learning to a company-wide shared intelligence layer. Its approach inverts the typical AI interaction: agents become proactive, recognizing when a skill applies and bringing it forward without being summoned.
- Artificial Intelligence
- AI Agents
- Enterprise Software
- Software Only
Funding
Funding not disclosed
Founders
Founder details are not available yet.
Product
Problem
Knowledge work is difficult to codify because what is "right" depends on the specific company, leaving organizations reliant on whether their workflows lie within a model's training distribution. Manual specification of workflows requires armies of forward-deployed engineers mapping companies interview by interview, and this cost is paid repeatedly as edge cases emerge and workflows evolve. A company cannot be specified once; it must be learned continuously.
Solution
Sonder is an applied AI lab building digital twins of organizational workflows, turning companies into continuous learning systems. The platform uses post-training vision models that live on a user's machine and watch their screen to learn from patterns, preferences, and workflows. This enables AI agents to observe work in motion, recognize when a skill applies, and act without explicit prompts—eliminating the need for human-initiated requests. By meeting people where they work, the system sees the workflow in real time and brings forward relevant capabilities without being summoned. This inversion shifts from people explaining their work to the work itself teaching the model, allowing work to move at the speed of thought. The immediate product helps individuals turn their work into reusable skills, which over time become shared team knowledge and eventually a company-wide intelligence layer.
Target Audience
Enterprises and organizations performing complex knowledge work that currently rely on forward-deployed engineering teams to codify workflows, particularly those with dynamic processes that change frequently. The product initially targets individual knowledge workers and scales to team and company-wide deployment.
Features
- Digital twin technology that models company workflows as they evolve in real time, built on vision models that watch a user's screen to capture cross-application context that APIs miss
- Promptless agent design where AI recognizes context and acts autonomously without user invocation, inverting the traditional interaction model
- Skills as living specifications: recurring work is turned into shareable skills that are refined by every use, correction, and edge case for the whole team
- Proactive capability delivery that surfaces relevant skills at the moment they apply, without being summoned
- Continuous learning mechanism that captures edge cases and workflow exceptions as they occur, from the individual up to the organization
- Human-centric approach that removes the bottleneck without removing the source of judgment, emphasizing synchronous collaboration between people and agents