Building the learning layer for enterprise safety.
Haven was built around a simple idea: every incident should make the organization smarter about the next one.
We combine enterprise AI, safety expertise, and company-specific knowledge to help high-risk organizations investigate more consistently, assure quality, and learn across every event.
Why Haven exists
Safety organizations generate enormous amounts of knowledge through incidents, investigations, corrective actions, procedures, and operating experience. Too much of that knowledge remains trapped in individual reports and systems.
Haven exists to make that knowledge usable. Not by replacing the people responsible for safety decisions, but by giving them a reasoning and learning layer that can connect evidence, history, controls, and organizational context at a scale no individual person can maintain alone.
Read Our Story →Built with industry and AI expertise
Haven was built at the intersection of enterprise safety and artificial intelligence, with the goal of solving real operational problems rather than adding generic AI to an existing workflow.
The company was co-founded in partnership with The AES Corporation (NYSE: AES), a global energy company operating in 12 countries, and AI Fund, the venture studio founded by Andrew Ng — bringing together state-of-the-art AI expertise with deep safety and field experience.
Our mission & values
Help high-risk organizations learn faster from what goes wrong, guided by precision, learning, trust, and operational relevance.
Mission & Values →Meet the people behind Haven
Enterprise software, AI, safety, public health, and high-risk operations experience.
Meet the Team →Every incident should leave the organization smarter.
That belief is the reason Haven exists.
The problem was never a lack of safety data.
High-risk organizations already generate large amounts of safety information: incident reports, witness accounts, photographs, corrective actions, procedures, standards, audits, investigations, and years of historical records.
The problem is that the knowledge inside that information is difficult to carry forward. Investigators still spend significant time assembling evidence and reconstructing events. Quality varies. Previous incidents are hard to search meaningfully. The same controls can fail repeatedly without the organization recognizing the pattern early enough.
We started with investigation because reasoning had to work first.
Investigation is where raw evidence becomes organizational understanding. Haven was designed to help investigators reason through complex events with greater structure, consistency, and efficiency while keeping accountable people in control of the conclusions.
That work created the foundation for something larger. Once the organization can reason more consistently about one incident, the same reasoning layer can begin connecting many incidents.
The larger opportunity is enterprise learning.
A completed investigation should not become a static report. It should strengthen the next investigation, inform the next corrective action, reveal recurring control failures, and help leaders understand where the organization continues to struggle.
That is the Haven learning loop: Investigate, Assure, Learn, Prevent.
Built for high-risk operations, not generic AI use cases.
Haven combines AI reasoning with company procedures, investigation standards, historical incidents, controls, corrective actions, and industry knowledge. The objective is not to make AI sound confident. It is to make safety reasoning more evidence-based, traceable, and useful in the operating context of the organization.
The future of safety AI is not automation for its own sake.
The opportunity is to help people see more, reason more consistently, and learn at enterprise scale while keeping human accountability where it belongs.
Haven is building that learning layer.