Six modules. One product.
Everything below ships together, with one login and one workflow engine, and every module drains into the same action register.
Near-miss reporting
A two-minute guided conversation instead of a form, so more of what happens gets reported, complete.
- Describe it in plain language; the system asks what an investigator would ask
- Photos, timestamp, and location captured automatically; any device, offline
- Classification suggested on capture, with reasoning shown and one-click override
- An investigation draft follows the report: system, human, and physical factors, direct cause, and root cause, with reasoning filled for review
- Coming soon Voice and phone-call reporting: describe it hands-free or call it in
Hydraulic hose weeping at press 2, small puddle under it
Is the press still running?
No, stopped it and tagged the isolation valve
Anyone exposed or injured?
No, area is clear, told my supervisor
Photo attached
One app, one form, from report to drafted investigation.
Unsafe act & unsafe condition reporting
The same two-minute capture for the daily, highest-volume signals.
- Unsafe acts and unsafe conditions, captured in the same conversational flow as near-misses
- Classified automatically on capture: unsafe act, unsafe condition, process safety or workplace safety near-miss, with one-click override
- The same drafted analysis as near-misses: contributing factors, direct cause, and root cause
- Routed to the right reviewer and into the action register; closed coaching-level or escalated by severity
The weak signals get reported while they are still cheap to fix.
Incident reporting & investigation
Formal incidents, from first report (FIR) to lessons released plant-wide (LFI). Draft-first, so investigations finish on time and miss less.
- FIR
Formal incident reporting
The event is reported formally, severity assessed, and an investigation sponsor and team assigned by severity.
- FAR
Full analysis report
Structured analysis through standard methods, evidence folded in, findings under dual approval before release.
- LFI
Lessons from incidents
Lessons prepared from the closed investigation and released plant-wide through the system, not by email.
- Severity assessed at the first report; serious events get a sponsored investigation team and dual review before release
- The full analysis report (FAR) starts from the AI-drafted analysis, in your report format (AI-assisted investigation below)
- Completion targets with reminders, so investigations keep moving
- An incident closes only when every action from it closes
Your formats, digitised: FIR, FAR, and LFI as your plant writes them.
Safety observation rounds
Structured rounds run as coaching, not policing.
- Checklist rounds on mobile: safe / unsafe / N.A. per item, your categories
- Unsafe items carry an injury-potential rating and the action taken, so there are no empty findings
- Coaching-level items close on the spot; higher severity routes to review
- Safe and unsafe behaviour trends by area, on the dashboard
-
PPE: gloves and eye protection in use
Safe -
Tools: right tool for the job, good condition
Safe -
Body position: line of fire, beneath suspended bag
Injury potential MediumAction taken Coached on the spot: repositioned, discussed safe standing zonephoto attached -
Reactions of people: adjusting work when observed
N.A.
Designed as a coaching process, not violation evidence capture.
CAPA & action tracking
One register for actions from every source. Owner, due date, evidence, or it escalates.
- Near-misses, unsafe acts and conditions, observations, investigations, audits, walkthroughs, and committee meetings, plus source categories your admin defines
- Every action carries an owner, a due date, and evidence to close. Or it escalates
- One action can go to multiple departments, with closure tracked department by department
- Aging alerts before items go overdue; register export and a source-by-department matrix for leadership review
- Audit Due in 9d
Restore fire-water ring main pressure test records
- Near-miss Due in 4d
Replace gasket and re-torque flange, T-105 transfer line
- Observation Due in 2d
Mark safe standing zones at bagging conveyor
- Committee Overdue 11d
Update permit-to-work display boards, all units
Every report tracked to closure.
Dashboards & insights
Repeat incidents, hotspots, and program health, visible without anyone compiling a report.
- Role-based views: executives, HSE ops, area managers, and sponsors each see their program, plant by plant
- Repeat patterns, hotspots, and aging items surface on their own
- Ask questions in plain language, like "which areas had the most repeated near-misses this quarter?"
- Audit-ready registers, investigation reports, and closure reports in one click
Spill-related events in Tank farm classified Tier 3 on night shift, Tier 2 elsewhere. Review suggested.
The repeat shows up before it becomes the serious one.
AI-assisted investigation
Every event arrives with a drafted analysis, near-misses and unsafe acts included. The team works a draft, not a blank page.
- More than one root-cause scenario proposed, each traceable to its evidence
- Standard methods run on the same evidence: 5-Why, Fishbone, Contributing Factors, Bow-Tie
- System, physical, and human factors analysed with the reasoning written out, alongside direct cause and root cause
- Ask plain-language questions mid-investigation and get answers grounded in public incident learnings, with sources shown
- The team challenges, edits, and completes the RCA on the draft
Could the gasket have passed inspection and still failed?
Possible. If the torque record is current, consider thermal cycling on the transfer line. Added as an alternative line under Hypothesis B.
based on 2 similar past events + public incident learningsWhat are the most likely causes of heat exchanger tube ruptures?
Most documented cases trace to tube-side corrosion, vibration-induced fatigue, or overpressure during transient operation. Inspection intervals are the common gap.
sources: public investigation reports · industry guidance- 1 Oil on walkway
- 2 Flange leak, T-105 transfer line
- 3 Gasket condition at last service?
- Torque sequence in SOP not followed
- Thermal cycling on transfer line
Everything stays a draft until the investigation team reviews and signs it.
HSE assistant
Safety knowledge your workers can actually reach: ask in plain language, get an answer in plain language, on mobile and web.
- Any worker can ask, any time: procedures, permits, precautions, what applies to the job in front of them
- Answers draw on your plant's procedures and public safety guidance, with the source shown
- No hunting through PDFs for the one paragraph that applies
- Separate from the investigation copilot; this one is for every worker
What do I need before grinding work near the transfer line?
That counts as hot work near a live line. Before you start:
- Hot work permit signed by the area owner
- Gas test before and during the job
- Fire watch posted through the work
Who approves the permit on night shift?
The on-duty shift in-charge for that area.
The workflow engine & notifications
The six modules run on one engine, and the engine runs your workflow: approval chains, roles, and routing configured to how your plant already works.
Your own approval workflows.
Add and change approval chains: who creates, reviews, approves, assigns, and closes, per plant and department. The application follows your company's workflow.
A notification at every handoff.
Assignment alerts, review requests, due-date reminders, aging alerts, and escalation when something sits too long.
Releases carry the report.
Approved reports and lessons go out through the system with the PDF attached, not by someone circulating files.
Closure needs evidence.
Nothing closes on a checkbox: actions close with evidence and remarks, and every step is on the audit trail.
How can AI draft an investigation for a plant it has never seen?
It has seen more of your plant than you might expect, and it never gets the last word.
- 01
It works from your plant's procedures and details.
Raven reads your plant's safety procedures, your event taxonomy, your report formats, and your role structure. A draft cites your SOPs and refers to your areas, not a vendor's generic versions. Plants that want deeper grounding can connect more over time, like past incident records; nothing depends on it.
- 02
It's grounded in decades of public incident learnings.
Long before it reads your first report, the AI is grounded in publicly available incident investigations and guidance from bodies such as the U.S. Chemical Safety Board (CSB), CCPS, OSHA, and OISD. The material a good investigator has spent a career studying is built into how it reasons about causes.
- 03
It proposes lines of analysis. You decide.
More than one plausible root-cause path, run through standard methods, each traceable to its evidence. The investigator challenges, refines, and picks the line worth working. Mid-investigation, the team can ask plain-language questions and get answers grounded in public incident learnings, with sources shown.
- 04
Everything stays a human-owned draft.
Every AI output in Raven Incident is a draft a person reviews: editable, auditable, never auto-final. Classification can be overridden in one click. Investigation findings are owned by the team that signs them.
awaiting investigator review
Lessons your plant already paid for.
- Similar past events, automatically. When a new report arrives, related events from your plant's history appear alongside it
- Repeats get flagged, not filed. The third similar event in ninety days reads very differently from three unrelated entries
- Equipment and area hotspots, visible without anyone building the analysis by hand
- Classification anomalies. When the same kind of event starts getting classified differently across shifts, the dashboard flags it
Flange leak, transfer line · Area 3
Flange seep, transfer bay
same equipment typeOil sheen below pipe rack
same area · similar causeDrip at pump P-22 coupling
same category, night shift3rd similar event in this area in 90 days
Built for how plants actually run.
Android, iOS, and web apps.
Mobile apps for frontline reporting, even offline. The web workspace for review, investigation, dashboards, and administration.
Our cloud or yours.
Runs as a cloud service on enterprise infrastructure (AWS/Azure), or deployed in your own cloud. On-premise deployment options available.
Roles & permissions.
Role-based access with configurable role groups: who can create, review, approve, assign, and close, defined per plant and department.
Single sign-on.
Single sign-on with your identity provider.
Serious events get a formal track.
Higher severity routes to a formal investigation with a sponsored team and dual review before release.
Multi-plant, multi-department.
One system across plants and departments, with department-wise views of what's reported, open, and closed. One login across units.
Your admin runs it.
Users, roles, taxonomy, sources, approval chains: changes your admin makes the same day, not a vendor ticket.
Audit trail throughout.
Every action attributable: who reported, who classified, who approved, who closed, and when.
Security.
SOC 2 Type 2 certified, encryption in transit and at rest, tenant isolation. Security & data →
Coming soon.
WhatsApp & voice capture
Report a near-miss the way your plant already communicates: a WhatsApp message or a voice note will become a structured report. No new app for contractors.
Multilingual capture
Hindi, regional languages, Arabic: frontline teams will report in the language they think in; the record stays consistent.
API RP 754 tier auto-mapping
Events will be classified to Tier 1–4 automatically, with plant-level Tier 3/4 leading-indicator rollups for RP 754 conformance reporting.
CAPA effectiveness checks
After closure, the system will watch for recurrence of similar events in the same area, and flag when an action didn't work.
Ready to close the loop on safety?
See how Raven Incident captures complete reports, drafts investigations, and tracks every action to closure at your plant.