Root cause findings that hold up under scrutiny.
A deterministic, evidence-grounded RCA platform built for reliability engineers and quality teams — accelerating investigations from hours to minutes while producing structured, auditable outputs every time.
|
Not a chatbot. Not a summarizer. Every finding is traced to a specific evidence source. Every inference is logged. If the evidence doesn’t support the conclusion, the engine says so.
13
Analysis Modules
0%
Hallucination Risk
Global 8D
Connected Workflow
IATF 16949
§10.2.3 aligned
Audit-ready report
Evidence Grounded
Every root cause and recommendation traces back to a specific uploaded evidence document. No unsourced assertions.
Deterministic Logic
Structured prompt architecture enforces the same analytical rigour on every case — not shaped by conversation history or model drift.
Audit-Ready Output
13 structured modules with full chain-of-evidence logging — readable by engineers, acceptable to auditors, actionable by operations.
Platform Capability
Built for industrial-grade investigations
Designed around the failure modes, evidence types, and reporting standards found in oil & gas, manufacturing, and process-critical industries.
13
Structured analysis modules per investigation
6M
Causal domains analysed in every report
0%
Hallucination tolerance — unsupported claims are flagged, not published
AI
Contradiction detection — conflicting evidence is surfaced, not averaged away
Process
The 5-D investigation workflow — define to deliver
The RCA AI Engine covers D1 through D3 — from problem definition and evidence upload through full AI analysis to structured review. D4 launches the Global 8D corrective response. D5 closes the loop with CAPA effectiveness verification.
D1
Define
Work order & evidence
D2
Diagnose
13-module analysis
D3
Design
Review & refine
D4
Deploy
Global 8D
corrective response
D5
Deliver
CAPA (coming soon)
Effectiveness verification
D1 – Define
Open a work order. Define the incident — asset, failure mode, business context. Upload every evidence document: maintenance records, DCS exports, inspection reports, witness statements.
D2 – Diagnose
The engine runs the full 13-module analysis — 5-Why, FMEA, FTA, Causal Inference, Knowledge Graph, RBD. Every module cross-checked against your evidence and confidence-scored. Under 5 minutes.
D3 – Design
Review the structured report. Evidence sources, confidence scores, and hallucination risk flags per module. Contradictions surfaced explicitly. The introspection layer identifies blind spots. You review — not rubber-stamp.
Analysis Modules
13 structured modules in every investigation
Each module is generated independently, cross-referenced against the evidence set, and scored for confidence and hallucination risk. Expand any module to learn more.
5-Whys
Traces the failure back through successive causal layers — stopping only when the chain reaches a systemic or organisational root, not a proximate symptom. Each “Why” is cross-checked against the evidence set and flagged if unsupported. Produces a structured causal chain rather than a narrative guess.
Fault Tree Analysis (FTA)
Constructs a top-down logical diagram from the failure event to its contributing causes, using AND/OR gate logic. Identifies single-point failures and combinatorial failure paths that 5-Whys linear analysis may miss. Particularly effective on multi-causal failures with concurrent conditions.
Failure Modes & Effects Analysis (FMEA)
Identifies failure modes relevant to the incident, assesses severity and detectability, and maps effects on equipment, process, and safety outcomes. Outputs are grounded in the specific failure context rather than generic equipment databases, making the scoring defensible for post-incident CAPA documentation.
6M Analysis (Ishikawa / Fishbone)
Systematically evaluates contributing factors across six causal domains: Man, Machine, Material, Method, Measurement, and Milieu (environment). Each domain is assessed independently from the evidence set, preventing the common bias of over-attributing failures to the most visible cause.
Timeline of Events
Reconstructs the sequence of events leading to the failure from timestamps across all evidence sources — DCS exports, operator logs, maintenance records, and inspection reports. Flags timestamp conflicts between sources. A reliable timeline is the foundation for distinguishing cause from coincidence.
Ontology Mapping
Maps the failure against structured industrial failure taxonomies — ISO 14224, OREDA, and domain-specific ontologies loaded into the knowledge base. Identifies where the failure pattern aligns with known failure classes, enabling comparison to industry experience databases and regulatory guidance.
Causal Inference Model
Applies structured causal reasoning to distinguish correlation from causation in the evidence. Evaluates counterfactual conditions — what would have needed to be different to prevent the failure — and weights contributing causes by causal strength rather than chronological prominence. Critical for multi-factor failures where no single cause is sufficient.
Knowledge Graph
Builds a structured entity-relationship map of assets, failure mechanisms, environmental conditions, and human factors mentioned across the evidence set. Surfaces connections that linear text analysis misses — for example, a maintenance deferral on one system that enabled a failure mode on an adjacent system.
Audit Notes
Generates a structured log of analytical decisions, evidence gaps, assumptions made, and rationale for root cause attribution. Designed to satisfy the documentation requirements of ISO 9001 corrective action procedures, IATF 16949, and AS9100 quality management systems — without requiring the analyst to write it manually.
KPI Analysis
Evaluates available process and maintenance KPIs against the failure context — MTBF, MTTR, OEE, and maintenance work order trends — to identify whether leading indicators were present before the failure event and whether they were acted upon. Supports both the RCA finding and the corrective action design.
Action-Item Log
Produces a prioritised corrective and preventive action list derived directly from the causal findings — not a generic checklist. Each action item is tagged to the causal layer it addresses, assigned a risk priority, and formatted for import into CMMS or quality management systems.
PM Analysis
Reviews the preventive maintenance programme against the failure mechanism — evaluating whether existing PM tasks were adequate in scope, frequency, and execution. Identifies gaps in the PM schedule that allowed the failure to develop undetected, and recommends specific task additions or frequency adjustments grounded in the causal evidence.
Reliability Block Diagram (RBD)
Models the reliability relationship between the failed component and its upstream and downstream system dependencies. Identifies whether the failure was isolated or whether the system design amplified its impact — informing both the corrective action scope and future design review recommendations.
Why This Plaform
What separates structured analysis from AI hype
Most AI tools in reliability are document summarisers with a quality disclaimer. This platform was built on a different premise.
Evidence traceability, not plausibility
Every causal claim is traced to a specific document and passage. If the evidence set doesn’t contain a supporting source, the module says so — it doesn’t infer or estimate.
Contradiction resolution, not averaging
When evidence sources conflict — operator log vs. DCS reading, maintenance record vs. inspection report — the engine surfaces the contradiction and applies structured logic to resolve it, with the decision logged.
Confidence scoring at module level
Each of the 13 modules carries its own confidence score and hallucination risk rating. Analysts see exactly which findings are well-supported and which require additional evidence before sign-off.
Knowledge base that compounds
Closed investigations feed a searchable knowledge base. New investigations are automatically cross-referenced against prior failure patterns — turning each completed RCA into an asset for future prevention.
Analyst in control at every gate
The engine produces structured outputs for review — it does not advance investigations autonomously. Every discipline gate requires human approval. Designed for quality systems that require a documented human decision chain.
On-premises deployment available
For clients with data residency requirements, the platform can be deployed with a local embedding model and vector store — no evidence data leaves the facility boundary. Cloud and hybrid configurations are also supported.
The Full Platform
Root Cause. Corrective Response. Verified Effectiveness.
The RCA AI Engine and Global 8D are both live. CAPA closes the loop — proving the fix actually worked, not just that it was implemented.
RCA AI Engine – Live
RCA AI Engine
13-module AI-assisted root cause analysis. Evidence-grounded. Contradiction-aware. Audit-ready.
- 13 analytical modules — 5-Why, FMEA, FTA, Causal Model, Knowledge Graph, RBD and more
- Evidence grounding with citation scoring — every finding traced to source
- Contradiction detection across all uploaded evidence documents
- Audit-ready PDF export — IATF 16949 and customer-ready
- Introspection layer — the AI audits its own analysis for blind spots
- Re-analysis cycle — reviewer feedback improves the report in place
Global 8D – Live
Global 8D
Complete D0–D8 structured corrective response. AI-assisted disciplines. Traceable to evidence.
- Complete D0–D8 discipline workflow with gates and status tracking
- AI-generated D2 problem description and D5 permanent corrective actions
- D7 lessons learned generation with knowledge base injection
- Full 8D report preview and PDF export — all disciplines, one document
- D8 sign-off gate — report locked until team formally closes
- Firestore persistence — resume across sessions, multi-user support
CAPA (coming Soon)
CAPA
Corrective and Preventive Action — effectiveness verified, loop closed, audit trail complete.
- Linked from RCA Cockpit and G8D D8 sign-off — no re-entry of investigation data
- AI-assisted preventive action proposals — grounded in confirmed root cause
- Effectiveness criteria with measurable targets and defined review dates
- Verification tracking — confirm the corrective fix held under real conditions
- Action register — owners, due dates, and closure status in one view
- Closes the 5-D loop — the Deliver phase confirmed, not assumed
Ready to run your first investigation?
The platform is available now. Open an incident, upload your evidence documents, and receive a structured 13-module RCA report in under five minutes.