After a year and a half of design, build, and field validation, today we are officially launching Haven Safety AI.
The AI-native safety intelligence platform was co-founded with and backed by The AES Corporation and AI Fund, the venture studio founded by the AI luminary Andrew Ng. Our mission is straightforward: help organizations investigate incidents faster, uncover systemic risk earlier, and prevent serious injuries before they occur.
You can read the full press release here.
Safety leaders do some of the most important work inside an enterprise. They are responsible for protecting people, earning trust with the workforce, and ensuring operations run reliably in environments where the stakes are high.
And yet, workplace safety is still one of the most technologically underserved functions in large organizations.
For years, EHS programs have steadily improved reporting, compliance, and baseline injury rates. But across many high risk industries, we have hit what many safety professionals recognize as the serious injury and fatality plateau. Severe outcomes persist despite continued investment in audits, training, and leading indicator programs. At the same time, workplace injuries cost the U.S. economy more than $160 billion each year.
That combination is why we built Haven Safety AI. Not to replace your current EHS, but to add more intelligence to one of the most critical workflows.
The missing capability in modern safety programs
Every serious incident leaves behind a trail of signals.
Those signals are real, and they matter. They show up in witness statements. In photos. In maintenance logs. In observations from the frontline. In near miss narratives. In prior incident reports that look similar, but live in another system, another site, or another folder.
The problem is not that organizations lack data. The problem is that the data is fragmented, mostly unstructured, and rarely connected in a way that supports fast, consistent, high quality decision-making.
As a result, safety teams …
- must manually reconstruct timelines from scattered inputs
- must do root cause analysis under time pressure
- must translate findings into corrective actions that hold up to scrutiny, including regulatory expectations
- must ensure actions actually reduce risk, not just satisfy a closeout checklist
- must do all of this while managing day-to-day operational demands
In many organizations, investigation quality varies widely by site and investigator. Lessons learned do not travel well. Institutional memory fades when people rotate. And the same classes of incidents repeat, sometimes with higher consequences.
We built Haven to change that dynamic.
What makes Haven different
Haven is an AI copilot for safety teams, designed to guide investigations and convert complex, unstructured information into clear, actionable intelligence.
From the start, we approached this as an AI-native product, not a traditional EHS system with an AI feature bolted on. The goal is not to replace safety professionals or their current EHS systems. The goal is to help them operate with more speed, more consistency, more insights, and more leverage.
Haven is designed to support the work that matters most:
- Guided, structured witness statements to reduce variability and missing details
- Automatic evidence gathering and organization across key inputs
- Multi-modal reasoning of images, voice, and documents
- Timeline synthesis that brings clarity to what happened and in what sequence
- Root cause identification that surfaces systemic contributors, not just proximate causes
- Corrective action recommendations grounded in regulatory standards and an organization’s historical context
- A growing institutional memory that improves learning and earlier intervention over time
In other words, Haven helps safety teams move from documenting what happened to understanding why it happened and what to change so it does not happen again.
A platform built for the full lifecycle of safety intelligence
Most tools stop at reporting. Real prevention requires a closed loop: capture, analyze, act, verify.
That is why Haven is organized into integrated modules that support the full lifecycle of safety management:
- havenSIGHT: Automatically collects witness statements, processes images, and captures frontline observations.
- havenEDGE: Analyzes incidents, surfaces causal patterns, and recommends corrective actions.
- havenIMPACT: Tracks outcomes over time and identifies leading indicators of future risk.
Together, these capabilities help transform safety from reactive reporting into proactive prevention.
If you saw our launch visuals, you also saw the concept reflected in the product experience: an AI-native platform for incident investigation and cause prediction, paired with prioritization of safety initiatives so teams can focus on the actions most likely to reduce risk.
Why now: AI is finally ready for safety’s hardest problems
Safety work generates vast amounts of complex information, and much of it is unstructured.
Historically, that reality limited what software could do. Traditional systems are good at forms, fields, and workflows. They struggle with the real substance of investigations: narratives, photos, nuance, context, and the web of contributing factors that sits behind severe outcomes.
Modern AI changes the ceiling on what is possible.
Haven combines artificial intelligence with a structured industry knowledge graph, enabling it to interpret and connect the kinds of information safety teams have always had, but have never been able to fully operationalize at scale.
This is why we can help teams:
- Move faster without sacrificing rigor
- Standardize investigation quality across sites
- Detect patterns that are easy to miss when information is siloed
- Learn from every incident and near miss, not just the ones that trigger major reviews
- Build toward predictive insights that support earlier intervention
Built with operators, validated in the field
Haven was co-founded in partnership with The AES Corporation, a global energy company operating in 12 countries. We also partnered with AI Fund, which brings deep experience building AI-first companies that solve real operational problems.
This combination matters.
High consequence operational environments demand practicality. The product has to work with real constraints: time, complexity, variability across sites, and the reality that safety teams are often asked to do more with less.
Over the past year and a half, we built, tested, and field validated Haven with those conditions in mind. The platform is now live.
AES has been clear about the goal: using every tool available to help ensure team members get home safely at the end of every workday. We share that commitment, and we built Haven to meet it.
Who we built Haven for
Haven is initially focused on industries with complex operations and elevated risk profiles, including:
- Energy and utilities
- Construction
- Manufacturing
- Logistics
If your organization operates in environments where a single incident can have serious human consequences, you understand the challenge: the difference between a “closed” investigation and a truly effective corrective action can be the difference between repeat exposure and lasting risk reduction.
Haven is built for that reality.
What to expect next
Launching is not the finish line. It is the beginning of expanding what safety teams can do when they are supported by AI that is designed for their work.
As we continue to grow, our priorities are consistent:
- Help safety teams execute faster and more consistently from day one
- Reduce repeat incidents by improving the quality and follow-through of corrective actions
- Build institutional memory that stays with the organization
- Enable earlier detection of systemic risk, before it becomes a serious incident
If you are leading safety, operations, or EHS transformation and want to see how Haven works, we would like to talk.