Complete capture
Guided evidence collection should reduce incomplete inspection returns.
Target hypothesis≥30% fewer incomplete returnsAILC · evidence before scale
See the case.
Challenge the idea.
ARIA is a product hypothesis for engineer-supervised pump DCIA—disassembly, cleaning, inspection, and assessment—not a finished autonomous assessor. This lifecycle makes the evidence, assumptions, alternatives, behaviours, and approval gates visible so partners can decide what deserves a pilot.
The AILC
Seven gates. One visible decision trail.
Each stage has an artifact, an owner, and a status. A later stage cannot silently turn an unvalidated hypothesis into an approved claim.
- 00ResearchDraft complete · field discovery open
- 01IdeaDraft · partner approval required
- 02BehaviourSeven epics ready for validation
- 03DevelopmentTarget design · not production proof
- 04IntegrationVerification gates defined
- 05ProductionNO-GO pending evidence
- 06FeedbackControlled learning loop defined
Research question
Can guided multimodal capture reduce DCIA rework without weakening engineering authority?
Faster review
Evidence-linked AI drafts should reduce reviewer effort without lowering quality.
Target hypothesis≥25% lower median review timeConsistent decisions
Image, measurement, and voice together should reduce clarification cycles.
TestCompare multimodal vs. image-only casesShop fit and scale
Capture must work in real conditions while data and policy remain isolated by shop.
GuardrailZero cross-shop leakageThese are discovery and pilot targets—not measured customer results. Baselines, representative jobs, validated hardware, Epicor fields, and approved taxonomy remain open evidence.
DCIQ business-case update
Standard work first.
Digital capture second. AI third.
DCIQ is the end-to-end disassembly-to-quote value stream. ARIA controls the DCIA technical record and approved estimating handoff; commercial quote release remains outside the product.
- 01 · BASELINE
4–6 weeks
Measure at least 20 comparable jobs where volume permits. Approve IDs, plans, shot lists, defect terms, tolerance sources, calibration, ownership, and exceptions.
- 02 · SHADOW
First 5 jobs
Use structured forms, QR identity, controlled photography, voice, and calibrated gauges with no downstream export. Verify association, record quality, and shop safety.
- 03 · PILOT
12–20 jobs · 8–12 weeks
Test completeness, documentation time, total DCIQ, corrections, revisions, adoption, and control compliance. Targets are hypotheses—not promises.
- 04 · SCALE GATE
Two stable months
Require reliability targets, positive verified ROI, stable adoption, and cross-functional approval before scanner purchase, AI expansion, or another shop.
BUILD FIRSTControlled capture cell + portable kit
Identity, required views, structured notes, measurement traceability, completeness, and approval.
DEFERAutonomous technical or commercial decisions
No photo-only crack or tight-tolerance acceptance, failure prediction, autonomous repair scope, estimate, or quote release.
Production remains NO-GO. A software test pass does not satisfy the operating, metrology, safety, adoption, ROI, or release gates.
Competitive context
Adjacent solutions validate the need—not this exact workflow.
Public vendor material reviewed on August 12, 2026 describes pump analytics and machine health. The current differentiation hypothesis is narrower: bench-level, multimodal DCIA evidence tied to an engineer-approved assessment and estimating handoff.
| Alternative | Public strength | Overlap | Gap versus the concept |
|---|---|---|---|
| Flowserve RedRaven | Connected diagnostics, trends, predictive monitoring | Pump-specific analysis | Public focus is operating assets, not teardown evidence |
| Sulzer BLUE BOX | Pump physics/AI, performance and remaining-life insight | AI plus pump expertise | Public focus is condition and performance data |
| KSB Guard | Vibration/temperature monitoring and maintenance tracking | Fleet visibility and API integration | Condition monitoring rather than component disposition |
| Augury | Cross-manufacturer AI diagnostics with expert validation | Human-supported AI at scale | Horizontal machine health, not DCIA bench workflow |
| Forms and spreadsheets | Familiar and flexible | Can record evidence and sign-off | Fragmented traceability and inconsistent completeness |
| ERP / EAM / CMMS | Work orders, approvals, purchasing, system of record | Job and quote context | Usually consumes findings instead of producing them |
Conclusion boundary: reviewed public materials do not describe the same end-to-end workflow. This is not proof that no private, custom, or unreleased competitor exists.
The idea
Tag → model → part → evidence → assessment → approved report.
The mobile workflow confirms the equipment tag and model first, proposes pump-part detections for technician correction, captures photographs, measurements, and time-coded engineer narration, then drafts an evidence-linked assessment and photo report for named technical review.
OCR, part detection, transcripts, and findings remain traceable candidates or drafts.
Identity correction, transcript confirmation, recapture, hold, approval, and attribution stay human-controlled.
ARIA hands off only an approved, versioned assessment report and reconciles delivery.
Observable behaviour
The product must prove the closed loop.
Govern shops
Provision isolated organizations, shops, memberships, policies, and audited access.
Confirm identity
Capture the equipment tag first; confirm model, pump type, assembly, and detected parts.
Capture field evidence
Require component views, measurements, calibration, and time-coded engineer narration with confirmed transcripts.
Draft assessments
Ground every observation in photographs, measurements, or audio timestamps and abstain when support is insufficient.
Review decisions
Compare, correct, recapture, hold, approve, and preserve history.
Publish safely
Version assessment reports with photo appendices and reconcile approved, idempotent estimating handoffs.
Learn under policy
Use corrections as controlled feedback without default shared-model training.
AILC evidence source
The pump taxonomy starts in the evidence folder.
SOURCE WORKBOOKpump_taxonomy comments EZ.xlsx
ailc/00-research/evidence/
The current idea artifacts use this component vocabulary to drive identification and B3 capture-plan design. It is a configurable starting point—not an approved universal taxonomy, damage criterion, tolerance set, or licensed standard.
CasingImpellerShaftShaft sleeveSeal chamberMechanical seal / packingBearingsBearing frameCouplingFasteners / gaskets
Identifies what is being inspected.
Drives views, dimensions, voice prompts, and references.
Records which taxonomy and criteria governed the inspection.
Engineering approves vocabulary, criteria, and disposition use.
Open work: engineering must validate hierarchy, aliases, damage modes, severity, measurement methods, tolerances, capture requirements, dispositions, and permitted standards references before pilot use.
Business partner validation
What should we believe, change, or stop?
Your response becomes decision input for the Idea gate. It does not approve production or replace engineering validation.
- Is the DCIA problem worth a controlled discovery?
- Is the differentiation specific and credible?
- Are the one-shop pilot targets decision-useful?
- What evidence is missing before investment?