A product by MindScript IT · In validation

Evidence Intelligence for Manufacturing

Turn scattered product documents into verified answers.

TraceFlow AI helps manufacturing teams structure information scattered across PDFs, emails, spreadsheets and supplier documents, connect important data points to their original evidence, and identify missing, inconsistent or outdated information.

Discuss a PilotTest the Workflow

Interactive prototype · fictional demo data · simulated extraction

  • Starts from a real client request
  • Every answer keeps its source
  • AI proposes. Humans verify.
TraceFlowS-042
Review in progress

Demo response readiness

3

verified

4

open items

7

requested

supplier_declaration_leatherworks.pdfPage 3
  • Upper material compositionFull-grain leather 78%, recycled polyester 22%Verified
  • Lining materialRecycled polyester 100%Verified
  • Sole materialNatural rubberSources disagree
  • Country of assemblyItaly — Civitanova MarcheVerified
  • Leather country of originMissing evidence
supplier_declaration_leatherworks.pdfPage 3

The upper is made of full-grain leather (78%) and recycled polyester textile (22%), supplied under contract L-18.

Version
v1.2
Uploaded
12 / 03 / 2026
Owner
LeatherWorks Italia
Awaiting review

Supplier action

LeatherWorks Italia

Leather origin · current REACH declaration

ENDraft ready

DPP preview — demo

S-042

Accepted answers only

Demo preview — not published

Demo preview — not scannable

The work behind the answer

The answer is the output. The searching is the problem.

A client asks for composition, origin, certificates and environmental information for one product. The same piece of information can sit in a bill of materials, a technical pack, a supplier declaration, an expired certificate and a production spreadsheet — each owned by a different person.

01

Collect

Someone chases documents across mailboxes, shared folders and colleagues, often depending on where a file happened to be saved.

02

Check

Dates, origins, compositions and declarations are read by hand, sometimes against sources that quietly disagree.

03

Repeat

The next client asks a similar question, and the same search starts again, because nothing durable was kept from the first one.

  • PDF
  • XLSX
  • EMAIL
  • CERTIFICATE
  • ERP

TraceFlow evidence layer

Structured fields · source kept · gaps identified

Client response

Digital Product Passports are creating a growing need for structured, accurate and traceable product information. TraceFlow focuses on the evidence layer behind that information.

Product thesis

The answer is the visible part. The workflow that finds, sources and verifies the evidence is the product.

TraceFlow AI is not a chatbot over your documents. It turns what it finds into structured fields, keeps each one attached to the document and location it came from, shows what is missing rather than filling it in, and requires a person to accept a value before it becomes an answer.

How it works

One repeatable path, from an incoming request to an evidence-backed answer.

Client Request → Structure → Evidence → Verify → Identify Gaps → Respond

  1. 01

    Start from the client request

    The request is turned into a checklist of fields. Everything downstream — readiness, gaps, the final answer — is measured against that list.

    TraceFlowS-042
    Review in progress

    Demo response readiness

    Response being assembled

    01 / 06
    clientRequest

    Giulia Ferrari · Meridian Sport Retail

    “Please provide the material composition, country of origin, certificates and sustainability information for AeroStep Runner, SKU S-042.”

    • Upper material compositionPending
    • Lining materialPending
    • Sole materialPending
    • Country of assemblyPending
    Fictional demonstration. The percentage is a ratio of verified required fields to total required fields for this request — not a compliance score and not a measure of accuracy.
  2. 02

    Load the sources and structure them

    Bills of materials, technical packs, declarations, certificates and production records are read into the requested fields. Source collection belongs here.

    TraceFlowS-042
    Review in progress

    Demo response readiness

    Response being assembled

    02 / 06

    Loaded sources

    bom_s-042.xlsx2 sheets · Materials, Notes
    tech_pack_s-042.pdf12 pages
    supplier_declaration_leatherworks.pdf4 pages · issued 12 March 2026
    reach_declaration_2023.pdf1 page · expired 31/12/2023
    production_record_s-042.xlsx1 sheet · Assembly
    Fictional demonstration. The percentage is a ratio of verified required fields to total required fields for this request — not a compliance score and not a measure of accuracy.
  3. 03

    Attach the evidence

    Each proposed value is shown with its document, the page or cell inside it, and the sentence that supports it. Where two sources disagree, both are shown.

    TraceFlowS-042
    Review in progress

    Demo response readiness

    Response being assembled

    03 / 06
    Upper material compositionFull-grain leather 78%, recycled polyester 22%Proposed by AI
    supplier_declaration_leatherworks.pdfPage 3

    The upper is made of full-grain leather (78%) and recycled polyester textile (22%), supplied under contract L-18.

    Version
    v1.2
    Owner
    LeatherWorks Italia

    Proposed

    Fictional demonstration. The percentage is a ratio of verified required fields to total required fields for this request — not a compliance score and not a measure of accuracy.
  4. 04

    Let a person decide

    The reviewer accepts, corrects or rejects each value and chooses between conflicting sources. Nothing becomes an answer without that step.

    TraceFlowS-042
    Review in progress

    Demo response readiness

    Response being assembled

    04 / 06
    Upper material compositionFull-grain leather 78%, recycled polyester 22%Verified
    Lining materialRecycled polyester 100%Verified
    Country of assemblyItaly — Civitanova MarcheVerified

    Decision

    Marco Rossi — Accepted without correction

    Fictional demonstration. The percentage is a ratio of verified required fields to total required fields for this request — not a compliance score and not a measure of accuracy.
  5. 05

    Identify and close the gaps

    Missing, expired and contradictory information becomes a targeted supplier follow-up. Whatever comes back re-enters human review.

    TraceFlowS-042
    Review in progress

    Demo response readiness

    Response being assembled

    05 / 06

    Gaps detected (4)

    Leather country of originMissing evidence

    No document in the loaded pack covers this field. The supplier declaration states on page 4 that it covers composition only.

    REACH declaration of conformityExpired

    This declaration is valid until 31/12/2023.

    Sustainability informationMissing evidence

    No document in the loaded pack supports a sustainability or recycled-content claim for this product. TraceFlow does not fill this in.

    Sole materialSources disagree

    Sole: natural rubber, vulcanised, 4 mm.

    Supplier follow-up · draft

    Missing product information — AeroStep S-042

    “For AeroStep Runner (S-042) we need the country of origin of the leather and a current REACH declaration of conformity. Please attach the document behind each answer.”

    Draft ready — sent manually by the user

    Fictional demonstration. The percentage is a ratio of verified required fields to total required fields for this request — not a compliance score and not a measure of accuracy.
  6. 06

    Prepare the client response

    A structured answer: each requested field, the value, its source, who approved it, and anything that is still outstanding.

    TraceFlowS-042
    Review in progress

    Demo response readiness

    Updated after human approval

    06 / 06

    Client response preview

    Upper material compositionFull-grain leather 78%, recycled polyester 22%Verified
    Lining materialRecycled polyester 100%Verified
    Sole materialNatural rubberVerified
    Country of assemblyItaly — Civitanova MarcheVerified

    Still outstanding

    Anything without accepted evidence stays visible in the response instead of being dropped to reach a round number.

    Client response ready — 7/7 requested items verified

    Fictional demonstration. The percentage is a ratio of verified required fields to total required fields for this request — not a compliance score and not a measure of accuracy.
Test the Workflow

Inside the workspace

Every field, its source, and where it stands with a reviewer.

A static view of the same record the interactive prototype works on.

TraceFlow
S-042AeroStep Runner

Demo response readiness

Verified required fields ÷ total required fields for the current client request.

3 verified4 to handle

Requested fields

Evidence source

supplier_declaration_leatherworks.pdf

The upper is made of full-grain leather (78%) and recycled polyester textile (22%), supplied under contract L-18.

Location
Page 3
Version
v1.2

Supplier follow-up queue

  • LeatherWorks ItaliaLeather country of originNo source
  • LeatherWorks ItaliaREACH declaration of conformityOut of date
  • LeatherWorks ItaliaSustainability informationNo source

Session history

  • Marco Rossi accepted Upper material composition10:24
  • Marco Rossi accepted Country of assembly10:19
  • Marco Rossi requested evidence for Leather country of origin09:52

Session history only — the prototype has no durable audit storage.

DPP preview — demo

S-042

Accepted answers only

Demo preview — not published

Demo preview — not scannable

Open the interactive prototype

Fictional demonstration data. Figures are derived from the fields shown, not from customer results.

  • Every value keeps the page or cell it came from
  • Gaps become a targeted supplier follow-up
  • Nothing unverified reaches the client response

Human in the loop

AI proposes. Humans verify.

TraceFlow keeps four things apart: what the extraction proposed, what a document actually supports, what a reviewer accepted, and what goes into the answer. A verified answer keeps its evidence; information that is missing stays visible as missing.

  1. 01

    Proposed by AI

    A structured value is extracted. It carries no authority yet.

  2. 02

    Backed by evidence

    The value stays linked to a document, a page or cell, and a version.

  3. 03

    Accepted by a reviewer

    A named person confirms or corrects the value. Only this step creates an answer.

  4. 04

    Included in the response

    Only accepted values, with their sources, go into the client response.

Sample field

Upper material composition

Full-grain leather 78%, recycled polyester 22%

Proposed by AI
  1. Proposed by AI
  2. Backed by evidence
  3. Accepted by a reviewer
  4. Included in the response

A value never moves straight from extraction to the client response. Every step has to be crossed explicitly, and editing a value or replacing its evidence sends it back for review.

What makes it different

An evidence layer built around one business workflow.

Five things a general-purpose document assistant is not built to do — described as what TraceFlow does, not as a comparison with any named product.

01

Structured fields

Answers live in named fields with a defined meaning, so the same question next quarter is an assembly job rather than a fresh search.

02

Source relationships

Each value stays attached to the document, page or cell behind it. A reader can check the answer instead of trusting it.

03

Gap detection

Missing, expired and contradictory information is raised as work to do, rather than smoothed over in a fluent paragraph.

04

Human validation

A named person accepts each value. New evidence, including a supplier's reply, always comes back for review before it counts.

05

Reusable responses

A verified field with its evidence can answer the next client, the next audit and the next system without being rebuilt.

Being built

Vertical schema

A data model and validation rules shaped around footwear product evidence.

Intended advantage

Corrected data

Reviewer decisions are the material that improves extraction rules over time.

Being built

Supplier graph

Relationships between suppliers, materials, certificates and products become reusable across products.

Strategy

Local distribution

Working closely with manufacturers, laboratories and industry bodies in one region.

These four areas describe an intended strategy. None is established: TraceFlow AI is a product entering validation.

First pilot

A deliberately small pilot, with the measures agreed in advance.

The first pilot stays narrow on purpose: one manufacturer, one product line, a controlled set of documents, and a workflow that already causes pain today.

1

manufacturer

1

product line

10

references

3–5

suppliers

30

days

Planned schedule

  1. Week 0

    Scope, data minimisation and a measurement of how the request is answered today

  2. Week 1

    Loading 30 to 100 documents and agreeing the field model

  3. Week 2

    Human review of the proposals and identification of gaps

  4. Week 3

    Supplier follow-up and handling of what comes back

  5. Week 4

    Completed responses, a before-and-after comparison, and the decision to continue or stop

Validation targets

Targets defined before the pilot — not results

  • ≥ 90% of proposed values accepted or accepted with a minor correction
  • ≥ 60% reduction in the time to answer one client request
  • 100% of accepted fields keep their source and location
  • ≥ 70% of identified gaps closed during the pilot
  • A paid extension, a letter of intent, or an equivalent commitment

These figures are success criteria set before a pilot. None has been measured, on real data or otherwise.

How it is built

A modular system with human approval at the centre.

01

Workspace

Companies, products, suppliers, documents and users, kept separate per customer.

02

Ingestion

PDF, DOCX, XLSX and email, with the source metadata preserved.

03

Evidence engine

Structured extraction, matching between sources, and detection of contradictions.

04

Review workflow

Final authority

Acceptance, correction, rejection, assignment and decision history.

05

Supplier workflow

Targeted follow-ups and tracking of the evidence expected back.

06

Response

A structured client answer, an evidence report, and an optional passport preview.

This describes the intended architecture. Only the parts you can exercise in the prototype are built today; see the section below for what is demonstrated and what is not.

Handling your business data

What is demonstrated, what is planned, and what depends on the deployment.

TraceFlow AI is a prototype. The list below separates what the prototype actually does from what a real deployment would need, so nothing here reads as a description of live security controls.

Source preservation

Demonstrated in the prototype

Every value keeps the document and location it came from, and the excerpt can be opened and checked.

Human approval before use

Demonstrated in the prototype

No value counts as an answer until a reviewer accepts it, including evidence returned by a supplier.

Audit history

Planned

The prototype keeps a session history in the browser tab and loses it on reload. Durable, tamper-evident audit storage is not built.

Controlled access and roles

Planned

The demo has a fixed reviewer profile and no sign-in. Authentication and role-based access are design work, not implemented behaviour.

Data minimisation

Planned

A pilot would take only the documents needed for the agreed product scope, and agree retention and deletion in writing beforehand.

Private or European deployment

Depends on the deployment

Running inside your own environment is possible in principle, and would need technical work and infrastructure agreed with your team. It is not available as a switch.

No training on your documents

Planned

A commitment we would put in the pilot agreement. The prototype processes nothing, so there is nothing to demonstrate yet.

None of this is a certification, and a reviewed declaration inside TraceFlow is evidence about a field — not a certificate issued by TraceFlow or by MindScript IT.

Start narrow

Footwear in the Marche region is the starting point.

A focused entry point: an industry with complex product documentation, where the same evidence problems repeat across products and suppliers, and where we can work closely with the people doing the job. It is where we intend to validate the workflow — not a claim of customers or completed pilots.

Expansion only makes sense once a second manufacturer can be onboarded without rebuilding the product.

  1. 01

    Footwear in the Marche region

    Starting point
  2. 02

    Manufacturers, suppliers and laboratories

  3. 03

    A reusable product evidence layer

  4. 04

    Other manufacturing sectors with the same problem

    Hypothesis

Let us answer one real client request together.

We are looking for a manufacturer, an industry partner or a support programme willing to test the workflow on a real product line, with a controlled set of documents and measures agreed in advance.

Discuss a PilotTest the Workflow
  • Limited scope
  • Minimised data
  • Measures agreed first
  • Human decision preserved

TraceFlow AI is a MindScript IT product in validation. This page describes a target workflow and an interactive prototype with fictional data — not a deployed product, a certification, or a guarantee of compliance.