Today, we're thrilled to share that Veris AI is launching with $8.5M in seed funding from an oversubscribed round led by Decibel Ventures and Acrew Capital, with participation from The House Fund, Idris Mokhtarzada, Ian Livingstone and Dorothy Chang among others. We’re grateful to have partners who deeply share our vision for the future of enterprise AI.
The era of agentic AI is here, but critical gaps remain. Most enterprises and developers still struggle to effectively train AI agents within the specific context of their applications. Other industries, like self-driving cars and robotics, have solved for this by investing extensively in simulated environments for training. AI agents need an equivalent.
Building genuinely autonomous AI systems requires a fundamentally different approach – one that leverages their strengths in reasoning and generalization while carefully managing their inherent limitations like stochasticity and robustness. Veris is leading this paradigm shift, empowering enterprises to fully harness the transformative potential of agentic AI.
As former academics with 25+ years' combined experience helping companies implement AI solutions, at startups like System, Workmate, KeyLead Health, and at large companies like Google, we've seen first hand the barriers that are holding back the adoption of AI agents. Together, we came to a shared realization that we need a new approach for training agents - one grounded in experience rather than pre-written prompts or static tests.
Although foundation models are becoming more intelligent by the day, their PhD-level general intelligence does not translate neatly into the domain-specific tasks that enterprise customers need. For example, a personal assistant agent must navigate complex calendaring scenarios, while a fintech chatbot is required to follow a strict set of regulatory requirements and avoid giving unauthorized financial advice.
They need to learn the job, not just be told what to do.
They can solve many problems by simply being "smart," but they still need to learn the specifics of the job over time. Today's AI agents need to learn the job, not just be told what to do.
We believe the next breakthrough in AI performance will come from allowing agents to learn through experience. As AI pioneers like David Silver and Richard Sutton highlighted in 'Welcome to the Era of Experience', true learning happens when agents can explore, experiment, and learn by doing. This is why we're building Veris: an end-to-end platform to supercharge your AI agents' performance using proven techniques like reinforcement learning, memorization, and targeted fine-tuning, all fueled by simulated experience.
The obvious bottleneck of this approach—especially in high-value and sensitive domains—is the availability of a safe environment and the risk of allowing an "untrained" agent to be deployed in production. After all, you would never allow a cashier in training to interact with your customers unsupervised.
Veris AI makes experiential learning safe and scalable. We train AI agents in high-fidelity simulations that accurately mirror their eventual production environments. These digital sandboxes allow agents to learn their roles thoroughly, learning common tasks and navigating challenging edge cases without real-world risk. Critically, these simulations are continuously updated by observing real-world agent performance, ensuring the "simulation gap" remains as narrow as possible.
Building and maintaining these dynamic environments is a significant undertaking, and we are deeply indebted to the researchers and open-source community pushing the boundaries of the agentic stack. We are committed to contributing back to this vibrant ecosystem.
Today, companies across manufacturing, fintech, and customer support are already leveraging Veris AI to train production-ready agents that deliver tangible value.
Follow our journey on LinkedIn and X (@veris_ai). Let's build the future of enterprise AI, together.
