Main site

Retail AI agent · Transparency statement

Supply-Chain Transparency Statement AI Agent

Trace every sentence in the statement back to the evidence and the date behind it, run the approval and publication clocks per entity, and hand the director the year-on-year delta before they sign.

Typical delivery
4–6 weeks
Deployment
Your stack
Director signs
Pre-approval
After launch
Agent Care

What this agent does

Evidences the statement, never signs it

In

  1. 01

    An entity crosses a reporting threshold, and a scope determination opens for it against each regime separately.

  2. 02

    An entity is in scope, and the criteria that regime's statement must address are pulled.

Reason

  1. 03

    An assertion is drafted, and the source, the date and the system behind it ride with it.

  2. 04

    An entity does none of what California asks, and the disclosure says so.

  3. 05

    An entity repeats last year's wording, and the change record marks it as repeated verbatim.

Decide

  1. 06

    An entity is UK-incorporated, and its type decides who approves and signs.

  2. 07

    A deadline falls in Australia, and the clock runs per entity.

Out

  1. 08

    An entity publishes, and the homepage link and the written-copy request are both tracked.

  2. 09

    Execute write actions only inside the approval boundaries agreed during implementation.

Product statement

The agent assembles the evidence; a director or equivalent approves and signs. It evidences what the group can show and refuses to assert what it cannot — a better document is not better conduct.

Example workflow

One statement year, scope to signature

  1. 1Group structure receivedEntity register, consolidated accounts, prior statements or scope worksheet
  2. 2Evidence gatheredPolicies, audits, grievance records, training logs and the source and date behind each
  3. 3Statement draftedCriterion coverage, per-assertion evidence, unevidenced items and confidence
  4. 4Controls appliedEvidence-binding checks, criterion-coverage checks, clock checks and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is approved at any of them — the agent is assembling, and the director's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the approving director to review.

Low confidence

Adds a counsel read first.

Director approval

The draft is held with its unevidenced assertions, its change record and the confidence.

Approve · Amend · Send to counsel

Approved — released to sign

  1. 6Statement records updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedCriterion coverage, unevidenced assertions, register acceptance and post-publication queries

Amendments

Director amendments are counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Signing the statement as a responsible person.
    • Putting the statement to the governing body.
    • Assessing the effectiveness of your own actions.
    • Deciding that an entity need not report this year.

  • Automation boundary

    Agent acts unaided

    • ✓Test each entity against the scope criteria each regime sets out.
    • ✓Carry each assertion with the evidence and date behind it.
    • ✓Run the approval, publication and copy-request clocks.
    • ✓Report the assertions still unevidenced, and what each one is missing.

    A director signs it; a correction after publication is republished and explained in public.

    • Deciding what a risk finding means for a supplier.
    • Lodging the statement on a public register.
    • Building an admissibility pack for a customs authority.
  • Changes to scope, criteria or approval rules.

Example output

One assertion, annotated

What the agent records is attached to the evidence it came from.

Statement output · single assertion

Illustrative example
Criterion
UK reporting area · training
What the evidence shows
Modern slavery training delivered to buying teams in the named operating companies
Last verified
Within this year
Source on file
Training platform log
Confidence
86%
Disposition
Held for the approving director

As received

Taken from the training log and the policy record as filed — nothing on this side is asserted by the agent.

Evidence used

  • Training platform log
  • Policy record on file
  • Operating-company list

Why this is held

Last year's statement made the same claim, and the evidence behind it is the same.

What the score decides

Below the configured threshold the draft adds a counsel read before it reaches the director.

Action

ApproveAmendSend to counsel

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

Every group entity

From the entity register

03Assembly

Build the statement from the evidence

Draw on the policies, audits and grievance records on file. The mining agent's minerals evidence stops upstream; the retailer is the downstream party whose director signs, and a risk finding here has to survive being written down in public.

01Approved path

The right body must approve

In the UK, s.54(6) ties both acts to the entity type — board and a director for a body corporate, members and a designated member for an LLP. Australia keeps them separate, and its register carries a 2026 guidance document for each.

02Human review

Pending is not in force

Directive (EU) 2026/470, in force 18 March 2026, raised the CSDDD thresholds, moved application to 26 July 2029 and deleted the Union-wide civil liability regime and the climate transition plan obligation. The UK Bill introduced on 30 June 2026 is not law as at 22 August 2026.

04

Build an evidence trail

The assertion, the evidence beneath it and the director who approved stay on the statement.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • Group and entity data

    • ERP
    • entity register
    • Statutory accounts
    • filings

  • Supplier and audit data

    • Sedex
    • EcoVadis
    • Elevate
    • Audit reports
    • CAPs

  • Policy and training

    • Learning systems
    • Policy and code libraries

Agent

Supply-chain transparency statements

  • Reads the entities
  • Assembles the evidence
  • Holds for the director

  • Grievance and case records

    • Speak-up lines
    • Case management

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six layers between the model and the statement

Six rings, outermost first. Whatever clears the innermost is named in the map beneath it.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeNarrow to evidence gathering when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, criterion-rule and regime-configuration changes.Track
L4TraceabilityRecord the evidence, its date, the unevidenced items and what the director saw.Record
L3Director approvalHold drafts for the named director; it governs release, not whether the sentence is true.Gate
L2Policy guardrailsTest each assertion against the evidence cited and the criteria that regime sets; a failure returns it.Restrict
L1Confidence thresholdsRoute low-confidence drafts to a counsel read before the director sees them.Require review
Model coreDraft assembled — criterion coverage, per-assertion evidence, unevidenced items and confidence
L1 – L2
Test whether an assertion may stand
L3
Puts the release in a director's hands
L4 – L5
Keep the assertion and the evidence behind it
L6
Narrows to evidence gathering when signals degrade

How Nestack evaluates it

Evaluate the whole assembly — not only the published statement.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the statement the public reads

  1. E1Final-output evaluationDid each assertion cite evidence that actually supports it?
  2. E2Step-level evaluationDid the agent use the right entities, criteria and regime configuration?
  3. E3Tool evaluationDid it read and write the correct entity and the correct criterion?
  4. E4Confidence calibrationDo low-confidence assertions actually attract more director amendments?
  5. E5Slice evaluationHow does performance change across entity types and regimes?
  6. E6Business outcomeHow many assertions needed an amendment or a correction after publication?

Floor — the outcome the director answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, set at the stage where each one first appears.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • HG-03

      Superseded scope record

      An entity's revenue position is read from last year.

    Stage gathers

    Entity register, evidence, policies and prior statements

  2. 02 · Reasoning

    2 modes

    • HG-04

      Criterion answered too widely

      An entity-level claim is drafted from one operating company.

    • HG-06

      Repetition read as evidence

      Last year's sentence is matched to last year's record.

    Stage proposes

    Criterion coverage, evidence, gaps and confidence

  3. 03 · Tool / write

    2 modes

    • HG-02

      Negative criterion dropped

      A criterion the group does nothing on is left unstated.

    • HG-05

      Approval body unrecorded

      The draft records approval without naming the body.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • HG-01

      Assertion without evidence

      A sentence in the draft cites nothing on file.

    Stage returns

    The statement the director approves and signs

  5. 05 · Change / Version

    1 mode

    • HG-07

      Silent criterion regression

      A model or rule change loosens what counts as evidenced.

    Stage tracks

    Model, prompt, criterion rules and regime config

  • Sev-1 · agent acts outside the boundary
  • Sev-2 · a wrong assertion reaches the draft
  • Sev-3 · source degrades, draft goes to review
See the complete Retail & E-commerce failure-mode catalogue

Affected slices

The group reads covered until you split it by entity

A group-level criterion-coverage figure can read complete while one entity type holds most of the unevidenced assertions. No published compliance rate exists to quote, so Nestack reports yours.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Newly in-scope subsidiaries6.2%3.4×Review
Part-owned joint ventures4.4%2.4×Review
Franchised and licensed entities3.4%1.8×Watch
Established parent entities2.0%1.1×Normal

Bar: unevidenced-assertion lift vs. established-parent baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

The loop closes on a test, not a fix

The loop shuts when the unevidenced assertion is a regression case. That suite is what the next statement lodged is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    The unevidenced-assertion rate rises in one entity type.

  2. 02

    Diagnose

    The statement signed by an officer the Act does not name is read back through the approval record until the cause narrows to one.

  3. 03

    Improve

    Any change goes out with a number, and the entities that caused it attached.

  4. 04

    Verify

    One statement case still red is enough to hold the release back.

  5. 05

    Learn

    One case joins the suite, one line joins the statement record.

  6. Learn → Detect

    The return edge. Every later detection starts from a suite this cycle lengthened.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, statement assembly, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Statement scope discovery and boundary definition.Week 1
  2. 02Entity register and evidence-source review.Week 1, Week 2
  3. 03Regime criteria and approval-route rule mapping.Week 1, Week 2
  4. 04Entity intake and evidence normalisation.Week 2
  5. 05Statement assembly and evidence binding.Week 2, Week 3
  6. 06Confidence scoring and unevidenced routing.Week 3
  7. 07Director approval workflow.Week 3, Week 4
  8. 08Evidence-store and record integration.Week 5
  9. 09Criterion and approval cases.Week 4, Week 5
  10. 10Guardrails and approval controls.Week 4
  11. 11Statement-trail instrumentation.Week 5, Week 6
  12. 12Deployment, documentation and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.

Capability✓ in scope · — not at this tierPilotOne entity, one regimeProductionProduction evidence systemsAdvancedMultiple regimes / entities
Introduced at Pilot
Assembly to your entities and rules✓✓✓
Director approval✓✓✓
Criterion-coverage baseline✓✓✓
Introduced at Production
Reporting by entity—✓✓
Approval workflow in your systems—✓✓
Approved write-back—✓✓
Group-structure integration—✓✓
Introduced at Advanced
Multi-regime criteria rules——✓
Multi-stage board approvals——✓
High entity count——✓
Multi-regime statement controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.
Separate from build
Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

  1. 01Your group structure and entity register

    Entity intake and evidence-source mapping

    Week 1

  2. 02Your last two published statements

    Coverage baseline, evidence binding and the change record

    Week 2

  3. 03Your regime scope and approval routes

    Regime criteria and approval-route rule mapping

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Entity register and evidence-source review, then integration setup

    Week 2

  5. 05Statements you would not want compared

    Approval cases and the evaluation suite

    Week 4

  6. 06What no statement may assert

    Confidence scoring, gap routing, guardrails and approval controls

    Week 3

  7. 07A named director or equivalent to approve

    Director approval workflow, then pilot and production validation

    Weeks 5–6

Nothing else is required

Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Each width is the time that phase really occupies, and week five accordingly carries two bands.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Statement scope discovery, regime mapping and the automation boundary

W2Entity and evidence integration and the coverage baseline

W3Statement assembly, clock logic and approval controls

W4Evaluation suite, negative-disclosure handling and failure-mode testing

W5Evidence-store integration, a pilot entity and targeted corrections

W6One statement cycle run under the approving director, then Agent Care handover

Reading the band
A band covers the weeks its work is actually named in. The week 5 overlap is real, not padding.
At the end of W6
Once the statement validates, Agent Care assumes the agent.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Retail AI agent

Build a transparency-statement agent around the evidence your director signs.

Show us your group structure, your evidence sources and who signs. Your company secretary already knows which board approved last year's statement — we hand them the evidence behind each sentence, and the year-on-year delta, before the director signs.

Typical delivery
4–6 weeks
Deployment
Your group structure
Director signs
Pre-approval
After launch
Agent Care available

Nestack Agents · Supply-chain transparency statements

AGT-RT-23 · Agent Care available after launch

Build a transparency-statement agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your statements

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.