Request a Demo
English
English Español Français
Frequently Asked Questions

Direct answers about Haven, AI, investigation, integration, and enterprise deployment.

Platform
Haven is the AI reasoning and learning layer for enterprise safety. It helps organizations investigate incidents, assure investigation quality, learn across historical data, identify recurring control failures, and turn those lessons into stronger prevention.
No. Haven is designed to work alongside existing EHS platforms. Your current system can remain the system of record for incident records, workflows, actions, approvals, and enterprise reporting. Haven adds reasoning, assurance, and learning capabilities.
Haven is designed for enterprise safety, EHS, operations, investigation, risk, and quality leaders, as well as the investigators and subject-matter experts responsible for complex incidents in high-risk operations.

You know that you need Haven when one or more of these questions represent a challenge for your team today:

  1. Can you tell which safety controls repeatedly fail across incidents and sites?
  2. How do you know whether corrective actions actually prevent recurrence?
  3. Does your organization learn from every incident, or only the most serious ones?
  4. Can you objectively measure whether investigation quality is improving?
  5. If the same control failed across three different sites, would your current system connect the dots?
  6. Can you quantify your CAPA debt and identify where overdue, weak, or unverified actions create the greatest risk?
  7. What percentage of your CAPAs rely on training and procedures versus engineering controls, substitution, or elimination?
  8. Does investigation quality remain consistent across sites and investigators, or does it depend on who conducts the investigation?
  9. Can investigators see relevant past incidents, failed controls, and previous CAPAs while an investigation is still underway?
  10. When leadership asks, “Where should we intervene this week?” can your incident data provide a clear answer?
Haven is designed for high-risk operating environments such as energy, utilities, construction, industrial services, and other organizations where investigation quality and organizational learning materially affect safety performance.
Most organizations begin with a focused workflow such as complex incident investigation, investigation quality assurance, or historical control-failure analysis. Request a demo and we will map the conversation to your current process and data environment.
AI Agents
A Haven AI agent is a specialized reasoning capability responsible for a defined part of the investigation and learning process. Each agent has a specific role, relevant evidence, access to approved knowledge, and a focused analytical objective. Together, the agents operate as a coordinated safety reasoning system rather than a collection of independent chatbots.
Complex incidents rarely involve a single issue. Equipment, controls, procedures, training, human factors, operating conditions, and prior organizational decisions may all contribute. Haven uses specialized agents so each dimension receives focused examination. The Investigation Lead then connects those perspectives into one evidence-based analysis for human review.
The Investigation Lead coordinates the process. It reviews the available information, engages the appropriate specialist agents, tracks open questions, and synthesizes their findings. All agents work from shared incident evidence, company knowledge, historical context, and the Haven Industry Knowledge Graph, which keeps their analysis connected rather than producing separate, disconnected answers.
No. Haven engages the agents relevant to the incident, evidence, and questions involved. A complex process safety event may require several engineering, controls, procedural, and regulatory perspectives. A more straightforward incident may need a smaller combination of agents. The approach adapts to the investigation rather than applying unnecessary analysis.
Haven's agents can reason using four connected sources of context: evidence from the current incident; company procedures, programs, standards, and control frameworks; historical incidents, corrective actions, and outcomes; and Haven's Industry Knowledge Graph. This grounding allows the agents to evaluate an incident within the context of both recognized safety knowledge and how the organization actually operates.
The Historian and other specialist agents retrieve relevant past incidents, causal patterns, control failures, corrective actions, and verification outcomes. They look beyond whether two incidents appear similar. Haven can identify incidents involving the same underlying control failure, even when they occurred at different sites, involved different activities, or were categorized differently.
Haven does not force the available information into a single confident answer. The agents identify missing evidence, conflicting accounts, unsupported assumptions, unresolved questions, and competing causal explanations. These issues remain visible so investigators can collect additional evidence or apply professional judgment before reaching a conclusion.
Yes. Haven is designed to keep findings connected to the evidence, source material, historical context, and reasoning that support them. Investigators and reviewers can examine the basis for an AI-inferred finding or AI-recommended action, then accept, revise, reject, or investigate it further.
No. Haven's agents strengthen professional judgment rather than replace it. They perform time-intensive work such as organizing evidence, retrieving relevant knowledge, testing causal reasoning, identifying gaps, and evaluating potential corrective actions. Accountable people remain responsible for final conclusions, root causes, corrective actions, escalations, and approvals.
In INVESTIGATE, the agents organize evidence, reconstruct timelines, explore causal pathways, examine failed controls, and develop corrective-action recommendations. In ASSURE, they evaluate completed investigations for evidence quality, causal coverage, guideline adherence, control analysis, and corrective-action strength. In LEARN, they reason across incidents to reveal recurring control failures, CAPA debt, repeated response patterns, investigation quality trends, and emerging enterprise signals.
Investigation
Investigation is one part of the Haven platform. INVESTIGATE supports evidence synthesis, timelines, causal reasoning, control analysis, corrective actions, and reporting. ASSURE evaluates investigation quality. LEARN reasons across historical incidents to identify recurring failures and enterprise learning signals.
No. Haven supports the investigator's reasoning process. It can help organize evidence, test assumptions, explore causal pathways, and identify gaps, but the accountable investigator remains responsible for the conclusions.
Haven can support familiar approaches such as 5 Whys, Fishbone analysis, and multi-threaded causal reasoning. The objective is not to force every organization into a single method, but to improve the depth and consistency of the reasoning process.
Yes. Haven can compare evidence across witness information, documents, images, records, and other investigation materials to surface conflicting information or unanswered questions for investigator review.
Haven helps connect corrective actions to identified causes and failed controls. It can support evaluation of control strength, Hierarchy of Controls alignment, ownership, and verification requirements. The objective is stronger corrective action, not simply more actions.

Haven generates recommended corrective actions based on the causal findings from the investigation, particularly the controls that failed, were missing, or proved inadequate.

The recommendations consider the investigation evidence, identified root and contributing causes, existing company procedures and standards, equipment and work context, and Haven's knowledge graph of hazards, controls, failure modes, and prior incidents. Where available, Haven can also use the organization's historical incident and corrective-action data to understand what has been tried before and whether similar controls have repeatedly failed.

Importantly, Haven evaluates recommendations against the Hierarchy of Controls, prioritizing stronger systemic controls such as elimination, substitution, and engineering controls over relying primarily on administrative actions or retraining.

The objective is not simply to generate a list of generic CAPAs. It is to produce corrective actions that are explicitly connected to the identified causes and control failures and that are more likely to prevent recurrence.

Haven generates root causes by reasoning across the full body of investigation evidence rather than relying on a single statement, checklist, or generic RCA template.

It first reconstructs the event, including the timeline, people, equipment, hazards, procedures, expected controls, and actual conditions. A key part of the analysis is determining which controls should have applied and how they performed. Haven identifies controls that were implemented successfully, implemented but failed or were ineffective, and applicable but not implemented at all.

It then evaluates the evidence around those control outcomes, identifies contradictions and gaps, and develops and tests causal pathways across areas such as equipment, procedures, training, supervision, work planning, human factors, and organizational systems.

Where available, Haven also uses the company's procedures, standards, knowledge graph, and historical incidents to determine whether the issue is isolated or part of a recurring pattern.

The result is a set of root and contributing causes explicitly tied to the evidence and identified control failures, rather than simply restating what happened or defaulting to conclusions such as "operator error."

ASSURE
ASSURE is Haven's investigation quality-assurance capability. It evaluates completed investigations against company standards and a structured quality framework, surfacing missing evidence, contradictions, weak causal coverage, corrective-action gaps, and other issues that may require review.
No. ASSURE is designed to help expert review scale. It can identify which investigations are likely to need attention and explain why, allowing experienced reviewers to focus their time where judgment is most valuable.
Yes. ASSURE can assess completed investigation packages, including investigations created outside Haven, provided the relevant investigation materials and company guidelines are available.

ASSURE can assess an RCA based on the final investigation output alone, but the depth and confidence of the assessment increases significantly when the supporting evidence is also available.

With only the RCA report, Haven can evaluate areas such as the quality of the causal analysis, whether conclusions are logically supported within the report, root-cause adequacy, and the strength and appropriateness of corrective actions.

When the complete investigation package is available, including witness statements, photos, procedures, training records, equipment information, timelines, and other evidence, ASSURE can go much further. It can identify missing evidence, contradictions, unsupported conclusions, gaps in witness coverage, deviations from company procedures, and whether the stated causes are actually supported by the underlying evidence.

So, ASSURE does not require all supporting documentation to provide value, but the complete investigation package enables a much more rigorous quality assessment.

LEARN
LEARN analyzes historical incident data to surface recurring control failures, repeated corrective-action patterns, emerging themes, and other signals across the enterprise. It turns incident history from a passive archive into an active source of prevention intelligence.
It is the ability to identify when the same safety control has failed across multiple historical incidents, even when those incidents otherwise appear unrelated. This helps the organization understand whether a current event reflects a persistent system weakness.
Yes. Historical incidents can become part of the Haven knowledge layer so investigators and leaders can reason across previous events, failed controls, corrective actions, and organizational responses.
Integration
Yes. Haven can be grounded in company-specific knowledge such as procedures, investigation guidelines, safety standards, control frameworks, and other organizational requirements.
Integration depends on the use case. Haven can work with EHS systems, document repositories, historical investigation data, company procedures, HR or training information, assets, contractor data, and other enterprise sources. Deployments can start with the minimum data required for value and expand over time.
Enterprise

Generic AI is designed to answer a broad range of questions. Haven is designed for enterprise safety reasoning. It combines company knowledge, historical incident data, controls, investigation standards, and specialized safety workflows so outputs are grounded in the organization's operational context.

Haven also relies on a coordinated team of specialized AI agents rather than a single model answering every question. Each agent is responsible for a defined part of the investigation, such as evidence, controls, procedures, or historical patterns, and an Investigation Lead synthesizes their work into one traceable, evidence-linked assessment. A generic assistant has no equivalent structure, which is why its answers are harder to ground, verify, and connect back to your organization's own evidence and standards.

No. Haven supports human decision-making. It helps organize evidence, reason across information, identify patterns, and surface issues, but accountable people remain responsible for investigation conclusions, corrective actions, and operational decisions.
Request a Demo