01
Collect
Teams ask several contacts for documents and often depend on personal folders or old email threads.
MindScript project • In validation
AI and product traceability
From scattered evidence to a traceable product.
TraceFlow AI turns bills of materials, technical packs, supplier certificates and spreadsheets into verifiable product data, linked to its sources and ready for validation.
30-day target pilot • One product line • Human validation
- Initial vertical: footwear
- Every field keeps its source
- No publication without validation
TraceFlowS-042Product evidence
Review in progress
Illustrative readiness
18
verified
2
corrections
4
unresolved
supplier_declaration.pdfPage 3
- Material compositionCuir pleine fleur 78 %Verified
- Upper supplierFornitore L-18Verified
- Leather originMissing evidence
- REACH declarationv1.2 · 2023Expired
- Repair instructionsPending
supplier_declaration.pdfPage 3
…la tomaia è realizzata in cuoio pieno fiore (78 %) fornito da Fornitore L-18…
- Version
- v1.2
- Uploaded
- 12 / 03 / 2026
- Owner
- Ufficio qualità
Awaiting review
Supplier action
Fornitore L-18
Updated declaration + origin evidence
ITDraft ready
Product passport
S-042
Approved data only
Not publishedPreview · not scannable
The real bottleneck
The passport is the output. The hard work comes before.
The same piece of product information can be spread across a bill of materials, a technical pack, a certificate, a supplier email and an ERP record. Before a reliable value can be published, someone still has to reconcile these sources, check their validity and hunt for what is missing.
02
Verify
Dates, origins, compositions and declarations are checked by hand, sometimes against differing definitions.
03
Reuse
The same data gets re-entered for the next client, product or system because no durable evidence layer exists.
- XLSX
- CERTIFICAT
- ERP
TraceFlow evidence layer
Structured fields · source kept · gaps identified
Product record
Europe is moving toward more structured product information. TraceFlow prepares and organises the necessary evidence without claiming to replace regulatory expertise, certification, or the company’s own responsible decision-making.
Product thesis
The passport is the result. The workflow that gathers and verifies the evidence is the product.
TraceFlow AI doesn’t just answer questions about documents. It turns the information it finds into verifiable fields, keeps the source behind every value, and coordinates the actions needed to close the gaps.
How it works
A repeatable path, from legacy files to a verified product record.
- 01
Gather the existing evidence
The user creates a product reference and imports its bills of materials, technical packs, certificates, declarations and spreadsheets.
TraceFlowS-042Product evidenceReview in progress01 / 05Illustrative readiness
Record being built
Imported documents
bom_S-042.xlsxFeuille 1 · 132 lignes · v4Identifiedtech_pack_S-042.pdf12 pages · v2Identifiedsupplier_declaration.pdf4 pages · v1.2Identifiedreach_declaration_2023.pdf1 page · v1.2Identifiedcare_guide_IT.docx3 pages · v1IdentifiedSynthetic demonstration. The percentages illustrate the workflow and are not a legal compliance score. - 02
Turn documents into proposed fields
The AI identifies materials, suppliers, origins, dates, certificates and product information according to a footwear-specific schema.
TraceFlowS-042Product evidenceReview in progress02 / 05Illustrative readiness
Record being built
supplier_declaration.pdf · Page 3
…la tomaia è realizzata in cuoio pieno fiore (78 %) fornito da Fornitore L-18…
Proposed by AIMaterial composition
Cuir pleine fleur 78 %
Footwear schema · confidence 0,92 · not validated
Synthetic demonstration. The percentages illustrate the workflow and are not a legal compliance score. - 03
Link every decision to its evidence
The reviewer accepts, corrects or rejects the proposed values. The human decision remains the final authority.
TraceFlowS-042Product evidenceReview in progress03 / 05Illustrative readiness
Record being built
Material composition
Cuir pleine fleur 78 %
VerifiedDecision
Accepted without correction
M. Rossi · 14 / 03 / 2026 · 10:24
supplier_declaration.pdfPage 3…la tomaia è realizzata in cuoio pieno fiore (78 %) fornito da Fornitore L-18…
- Version
- v1.2
- Uploaded
- 12 / 03 / 2026
- Owner
- Ufficio qualità
Synthetic demonstration. The percentages illustrate the workflow and are not a legal compliance score. - 04
Identify and close the gaps
TraceFlow detects missing, expired or contradictory evidence, then prepares a precise request for the relevant supplier.
TraceFlowS-042Product evidenceReview in progress04 / 05Illustrative readiness
Record being built
Gaps detected
- REACH declarationValidity expired on 12/31/2023Expired
- Leather originNo document covers this fieldMissing evidence
Supplier request · draft
ITFornitore L-18
« Gentile fornitore, per il prodotto S-042 chiediamo la dichiarazione REACH aggiornata e la prova di origine del cuoio. »
Draft ready · sent manually by the user
Synthetic demonstration. The percentages illustrate the workflow and are not a legal compliance score. - 05
Prepare a reusable product record
Once fields are approved, TraceFlow exports a structured record and generates a preview of the product passport.
TraceFlowS-042Product evidenceReview in progress05 / 05Illustrative readiness
Updated after human approval
Completeness raised to 91% after human approval
Product passport preview
S-042
Approved data only
Not publishedPreview · not scannable
Excluded from the published view
Leather origin and REACH declaration stay out of the export until their evidence is provided and accepted.
Synthetic demonstration. The percentages illustrate the workflow and are not a legal compliance score.
TraceFlowS-042Product evidence
Review in progress
Illustrative readiness
Record being built
Imported documents
bom_S-042.xlsxFeuille 1 · 132 lignes · v4Identified
tech_pack_S-042.pdf12 pages · v2Identified
supplier_declaration.pdf4 pages · v1.2Identified
reach_declaration_2023.pdf1 page · v1.2Identified
care_guide_IT.docx3 pages · v1Identified
Product experience
The information that matters becomes visible within seconds.
TraceFlow
Illustrative readiness
24 fields tracked · 4 need action
18 verified4 to handle
Product fields
Evidence source
supplier_declaration.pdf
…tomaia in cuoio pieno fiore (78 %)…
- Location
- Page 3
- Version
- v1.2
Supplier action queue
- Fornitore L-18Leather originDraft
- Conceria M-04REACH up to datePending
Review history
- M. Rossi accepted Material composition10:24
- M. Rossi corrected Upper supplier10:19
- A. Bianchi requested origin evidence09:52
Passport preview
S-042
Approved data only
Not publishedPreview · not scannable
Synthetic demonstration. The percentages illustrate the workflow and are not a legal compliance score.
- Every value keeps its source page or cell
- Gaps become assignable actions
- Unvalidated data stays out of publication
Human in the loop
Trust doesn’t rest on a single AI score.
TraceFlow clearly separates what the AI proposes, what the documents support, what a reviewer accepts, and what can ultimately be published.
- 01
Proposed by AI
A structured value is extracted, but it doesn’t yet carry any authority.
- 02
Backed by evidence
The value stays linked to a document, a page, a cell and a version.
- 03
Accepted by a reviewer
An identified person confirms or corrects the value.
- 04
Published
Only approved information can appear in the export or the passport.
Sample field
Material composition
Cuir pleine fleur 78 %
Proposed by AI
- Proposed by AI
- Backed by evidence
- Accepted by a reviewer
- Published
A value never moves directly from extraction to publication. Every step must be crossed explicitly.
Not a document chatbot
An evidence infrastructure built around a business workflow.
General document AI
Searches and summarises documents
TraceFlow AI
Turns evidence into product fields
General document AI
Returns an answer with citations
TraceFlow AI
Keeps a source at the level of every field
General document AI
Answers a one-off request
TraceFlow AI
Maintains a product record over time
General document AI
Flags what it found
TraceFlow AI
Also identifies what is missing or contradictory
General document AI
Generic behaviour
TraceFlow AI
Initial schema centred on footwear
General document AI
Leaves follow-up outside the system
TraceFlow AI
Prepares and tracks supplier requests
Layer being built
Vertical schema
A data model and validation rules adapted to footwear-industry evidence.
Advantage being pursued
Corrected data
Review decisions progressively improve the extraction models and rules.
Layer being built
Supplier graph
Relationships between suppliers, materials, certificates and products become reusable.
Differentiation strategy
Local distribution
Partnerships with manufacturers, labs, consultants and industry bodies create market access.
These four areas describe the intended differentiation strategy. None are established yet: TraceFlow AI is a product concept entering its validation phase.
First pilot
Prove the accuracy, the time saved, and the will to continue.
The first pilot must stay deliberately limited, so as to measure value on a real workflow without forcing the replacement of an ERP or PLM.
1
manufacturer
1
product line
10
references
3 à 5
suppliers
30
days
Planned schedule
Week 0
Security, data minimisation and measuring the current process
Week 1
Import of 30 to 100 documents and validation of the field model
Week 2
Human review of extractions and identification of gaps
Week 3
Supplier requests and follow-up on responses
Week 4
Approved records, before/after comparison and the extension decision
Validation targets
Pilot targets — not achieved results
- ≥ 90% acceptance rate on extracted fields
- ≥ 60% reduction in preparation time
- 100% of accepted fields keep their source and version
- ≥ 70% of requested gaps closed during the pilot
- 1 paid extension, letter of intent, or equivalent commitment
These figures are the success criteria defined before the pilot. None has yet been measured on real data.
Trust architecture
A modular AI, with human approval at its centre.
01
Workspace
Companies, products, suppliers, documents, users and workspace isolation.
02
Ingestion
PDF, DOCX, XLSX, emails and OCR, with source metadata preserved.
03
Evidence engine
Structured extraction, entity matching and contradiction detection.
04
Review workflow
Final authorityAcceptance, correction, rejection, assignment and decision history.
05
Supplier workflow
Targeted requests, submission links and tracking of expected evidence.
06
Export
Structured record, evidence report and interoperable QR preview.
Trust requirements
TraceFlow AI is a product concept in validation. None of the points below is implemented today. Each is shown with its real status: a retained principle, an element planned for the pilot, or a condition to meet before any real document is processed.
- European hosting or a credible private deployment optionArchitecture principle
- Encryption in transit and at restRequirement before real data
- Role-based accessPlanned for the pilot
- History and audit logPlanned for the pilot
- Retention and deletion controlsRequirement before real data
- No training on client documents without explicit agreementRequirement before real data
- Human authorisation before any external sharingArchitecture principle
Start narrow, expand with proof
Footwear in the Marche region is the entry point.
TraceFlow starts with a precise workflow, an accessible manufacturer, and an industry where the same evidence challenges repeat across products and suppliers.
Expansion only begins once a second client can be onboarded without rebuilding the product from scratch.
- 01
Footwear in the Marche region
Entry point - 02
Manufacturers + suppliers + labs
- 03
Reusable product workspace
- 04
Textiles, furniture and other European industries
Hypothesis
Let’s build the first verifiable product record.
We’re looking for a manufacturer, an industry partner or a support programme ready to test the workflow on a real product line and a controlled scope.
- Limited scope
- Minimised data
- Metrics defined before the test
- Human decision preserved
TraceFlow AI is a MindScript project in validation. The elements shown on this page describe a target workflow and synthetic demonstrations, not a deployed product or a guarantee of regulatory compliance.