PartsCleanse AI | AI2COE Solution

Business problem

MRO Catalog Diagnostic.

MRO catalog deduplication, field quality, UOM consistency, and duplicate capital exposure.

The page is not a static brochure. The engine has a working upload, preview, map, validate, analyze, evidence, score, report, action, review, and score-history path inside the Industrial IQ portal.

Buyer ownerCFO, CIO, Procurement, Maintenance, and Materials leaders

Trigger eventERP migration, AI readiness, working-capital pressure, procurement leakage, reliability risk, audit readiness, or recurring operational review.

Input data requiredDescription

Upload workflowUpload CSV, preview fields, confirm AI-recommended mappings, validate required coverage, normalize where needed, then run the engine.

Diagnostic logicMap source columns, normalize values, run deterministic engine checks, then attach evidence and confidence before report output.

Sample evidence tablePublic sample mode exposes mapped evidence rows before the customer uploads private data.

Score outputCatalog health score: lower values mean higher duplicate, field-quality, UOM, supplier-alias, and obsolete-record exposure.

Executive report previewSample report pages show the decision narrative, evidence rows, confidence tiers, assumptions, limitations, PDF export, and next actions before private upload.

Report outputPartsCleanse AI Catalog Diagnostic Report with HTML, CSV evidence, PDF, and report email status for authenticated runs.

Governance controlsNo ERP write-back, no autonomous supplier outreach, confidence-tier labels, human-review routing, audit metadata, and clear sample-versus-uploaded-data labeling.

Upload workflow

Upload -> Validate -> Analyze -> Evidence -> Score -> Report -> Action -> Repeat.

Step	Layer	Customer experience
1	Upload	CSV export enters the parser. Source file retention rules are disclosed.
2	Map	ERP/CMMS aliases are inferred, then corrected or confirmed by the user.
3	Validate	Required fields, completeness, missing values, and confidence reducers are shown before run.
4	Analyze	Engine-specific analyzers generate findings, evidence, and impact estimates.
5	Govern	Findings receive confidence tiers and human-review status before any action.
6	Report	Executive report, evidence table, action tracker, and score snapshot are produced.

Public proof before account creation

Customers can inspect samples before uploading their own data.

Every sample result is labeled as sample mode. It demonstrates mapping, validation, scoring, evidence, confidence, report structure, action logic, and recurring-review context without claiming customer-specific exposure.

Run Industrial IQ Snapshot Run sample diagnostic Download sample PDF Sample CSV Mapping template Data dictionary Product page

Required data and field dictionary

Accepted fields, aliases, and mapping expectations.

Input	Need	Common aliases	Meaning
Description	Yes	description; item_description; material_description; maktx; short_text; part_description; long_text; desc	Item, part, asset, work-order, finding, or source-record description used by the engine.
Material Id	Recommended	material; material_id; material_number; matnr; item; item_number; item_id; sku; part; part_number; stock_code	Unique material, SKU, item, or spare-part identifier from the source system.
Manufacturer	Recommended	manufacturer; mfr; maker; brand; oem; oem_name	Manufacturer, OEM, brand, or maker name.
Mpn	Recommended	mpn; manufacturer_part_number; mfrpn; part_no; oem_part_number; vendor_part	Manufacturer part number, OEM reference, supplier part number, or equivalent identifier.
Supplier	Recommended	supplier; vendor; vendor_name; lifnr; supplier_name	Supplier, vendor, vendor alias, or purchasing source.
Uom	Recommended	uom; unit; unit_of_measure; meins; base_uom	Unit of measure such as EA, Each, PCS, Meter, MM, Inch, Set, or Pack.
Quantity	Recommended	quantity; qty; stock_qty; on_hand; qty_on_hand; unrestricted; labst; stock_on_hand	Quantity, balance, order quantity, stock quantity, or demand quantity depending on engine.
Unit Cost	Recommended	unit_cost; cost; price; moving_average_price; map; valuation_price; standard_price; unit_price	Unit cost, average cost, standard price, last purchase price, or valuation rate.
Site	Recommended	site; plant; werks; location; storeroom; warehouse; depot; facility	Plant, site, warehouse, storeroom, region, location, or operating unit.
Currency	Recommended	currency; currency_code; waers; iso_currency	Currency code for cost, value, price, or exposure calculations.

Recommended diagnostic data pack

What to upload for a stronger run.

Recommended file	Fields that improve confidence
Material or item master CSV	description, manufacturer, MPN, supplier, UOM, site, value
Inventory balance export	quantity, stock value, site, currency
Supplier / OEM reference file	supplier aliases, manufacturer names, OEM part numbers

Business impact model

Why this matters to the buyer committee.

Capital exposure

Capital exposure model

Duplicate-family value, confidence-adjusted exposure, recoverable working-capital range.

Operational risk

Operational risk model

False stockout signals, duplicate item creation, maintenance uncertainty, ERP migration blockers.

Decision output

Decision output model

Duplicate families, confidence tiers, review priority, clean CSV baseline, executive report.

Evidence and confidence

What the report proves.

Output layer	Example	Why it matters
Score	Catalog health score	0-100 signal with risk level and trend-ready snapshot.
Score formula	Deterministic calculation	The report exposes the scoring formula and component inputs; random scores are not used.
Duplicate family	Part family evidence	Description, manufacturer, MPN, UOM, size/material discriminator, and exposure.
Finding	PartsCleanse AI Catalog Diagnostic Report	Issue title, severity, source engine, and owner-facing action.
Evidence	Mapped source records	Source-row references, relevant fields, analyzer reason codes, and confidence tier.
Evidence graph	Source -> finding -> evidence -> action	The result carries an evidence graph for review, report, action, and score-history continuity.
Confidence	High / Medium / Needs Review	Coverage, completeness, source-field quality, and analyzer agreement.
Action	Owner review item	Recommended action, priority, due window, and review status.
Renewal value	Recurring management view	The report shows exposure identified, review queue size, actions created, and next review cadence.

Buyer interpretation

One diagnostic, multiple executive decisions.

Buyer	Decision question	Evidence source
CFO	Can the finding be tied to capital exposure, carrying cost, leakage, or payback discipline?	PartsCleanse AI
COO	Does the evidence reduce operating risk, downtime exposure, site friction, or service disruption?	PartsCleanse AI
CIO / ERP owner	Are source fields mapped, export quality visible, and ERP write-back avoided unless governed?	PartsCleanse AI
Procurement	Does the diagnostic expose supplier, PO, duplicate spend, stocked-but-purchased, or price-variance risk?	PartsCleanse AI
Maintenance / Reliability	Does the evidence affect work-order readiness, false stockout, shutdown coverage, or critical-spare confidence?	PartsCleanse AI
Data governance	Can findings be reviewed, accepted, rejected, audited, and defended after the report is shared?	PartsCleanse AI

Trust controls

This engine is governed before operational action.

Trust control

No ERP write-back

Findings are recommendations and review items, not autonomous ERP changes.

Trust control

Evidence traceability

Every finding links back to source records, mapped fields, and reason codes.

Trust control

Confidence tiers

High, medium, low, and needs-review logic is visible before action.

Trust control

False-positive control

Unsafe lookalikes are separated from high-confidence findings.

Trust control

Human review

Owners can accept, reject, defer, assign, or request more data.

Trust control

Audit trail

Report exports, reviewer actions, and status changes remain traceable.

FAQ

Questions buyers ask before running PartsCleanse AI.

What data does PartsCleanse AI need?

PartsCleanse AI requires Description. Optional fields such as Material Id, Manufacturer, Mpn, Supplier, Uom, Quantity improve confidence and business-impact precision.

What does PartsCleanse AI produce?

It produces PartsCleanse AI Catalog Diagnostic Report, a 0-100 catalog health score, evidence records, confidence tiers, recommended actions, and a review-ready executive summary.

Does AI2COE write back to SAP, Maximo, Oracle, or any CMMS?

No. AI2COE diagnostics are decision-support outputs. They do not change ERP, EAM, CMMS, procurement, inventory, or asset records automatically.

How does confidence tiering work?

Findings are ranked by source-field coverage, data completeness, evidence quality, analyzer agreement, and whether a human owner should review the recommendation before action.

How should leadership use the report?

Use the report to decide whether the issue is measurable, material, governable, and worth funding before starting a larger ERP, inventory, procurement, maintenance, or AI transformation program.