Main site

Electronics AI agent · repair obligation

Repair Obligation & Spare-Parts AI Agent

Measure each part order against the window Reg. (EU) 2023/1670 owed on the day it arrived, five working days and later ten, and leave the price and the impossibility call to a named person.

Typical delivery
4–6 weeks
Deployment
Your stack
Named approver
Measured only
After launch
Agent Care

What this agent does

Measures the window, never the price

In

  1. 01

    A part is ordered, and the window runs from receipt, inside a seven-year period counted from end of placement.

  2. 02

    A part is ordered in year six, and ten working days is owed; the first five years of that period owed five.

Reason

  1. 03

    A part is ordered by an end-user, and Reg. (EU) 2023/1670 sends only some part classes that far.

  2. 04

    A passport is demanded in a tender, and no Digital Product Passport obligation binds an electronics product today.

  3. 05

    Placement of the last unit ends, and operating-system and security updates are owed free for at least five years.

Decide

  1. 06

    A source-code release lands, and a security update is owed in four months, a functionality update in six.

  2. 07

    An order lands on a chain with no EU manufacturer, and Art. 5(3) walks the duty down to the seller.

Out

  1. 08

    A repair information form is voluntary under Art. 4(1); once given, its price holds thirty calendar days.

  2. 09

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

Product statement

The agent holds the availability file and measures delivery against the owed window; a named person decides a repair is impossible, approves the price and signs the declaration.

Example workflow

One order, receipt to close

  1. 1Part order receivedEnd-user request, professional repairer order or a service-network draw
  2. 2Availability file matchedModel, end-of-placement date, part class, audience and the window owed on the day of receipt
  3. 3Window recordedFive or ten working days, with the registration and access windows a professional repairer is owed
  4. 4Controls appliedClock checks, part-class checks, cascade checks and completeness confidence

No human action required

Stages 1 to 4 run unaided, and nothing is promised or priced at any of them — the agent is measuring, and the compliance lane opens at the completeness gate.

5DecisionBranches at the completeness gate
Window met

Goes to the product compliance lead to close.

Anything late or thin

Adds a regulatory counsel read first.

Product compliance review

The file is held with its clocks, its part classes and its gaps.

Close the order · Append evidence · Send to counsel

Closed — by a named person

  1. 6Parts and service records updatedOnly where write access and disclosure policy allow it
  2. 7Outcome evaluatedWindow margin, evidence completeness, counsel corrections and what a later review found

Corrections

Each counsel correction is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Deciding that a repair is impossible.
    • Approving a repair price, or any part of one.
    • Issuing or standing behind a repair information form.
    • Declaring a model conformant in EPREL.

  • Automation boundary

    Agent acts unaided

    • ✓Hold one availability file for each model and for each part class.
    • ✓Run the owed window from receipt of the order.
    • ✓Measure each actual delivery against the working-day window owed.
    • ✓Flag the day the five-day window turns into ten.

    Nothing is promised or priced except by a named person, inside the agreed boundaries.

    • Judging which party the Art. 5(3) cascade lands on.
    • Telling a customer a part class is out of scope.
    • Setting an operating-system support end date.
  • Changes to window rules, part classes or file fields.

Example output

One order, annotated

Directive (EU) 2024/1799 applied on 31 July 2026, and this file is what a person closes the order from.

Availability file · single order

Illustrative example
Order
Display assembly, slate tablet
Owed as
Display assembly to an end-user, year six of seven
Clock
Day 7 of ten working days
Window of record
Reg. (EU) 2023/1670, Annex II B.1.1
Confidence
Held unclosed
Held for
The product compliance lead, in person

As received

Taken from the availability file and the order receipt — Art. 5 reaches only the Annex II product groups.

What the file holds

  • The order receipt
  • End-of-placement date
  • Part class and audience

Why no price call here

A reasonable price is the Art. 5(2) test, and the Annex I form carries no signature block.

What the score decides

Below the configured threshold the file picks up a counsel read before.

Action

Close the orderAppend evidenceSend 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 order

From the parts system

03Evidence

Statutory window, not service level

Warranty triage and field service own commercial repair; this is the window the rule owed, and take-back sits with our EEE producer agent.

01Approved path

The clock changes gear

Five working days for the first five years of the availability period, ten for the remaining two — Reg. (EU) 2023/1670, applying from 20 June 2025.

02Human review

Where the score and the seal live

The repairability class sits in energy-labelling Reg. (EU) 2023/1669, and an EPREL verification taken without a trade-register code expires 22 April 2027.

04

Build an evidence trail

The part, the window it was owed in and the person who approved the repair stay on the record.

Integrations

Typical integrations

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

  • Parts and service systems

    • ERP parts master
    • spares
    • Service-network order feeds

  • Product and lifecycle records

    • PLM
    • model masters
    • End-of-placement records

  • Registration and labelling

    • EPREL model registrations
    • Repairer registration portal

Agent

Repair obligation & parts

  • Reads the order
  • Measures the window
  • Holds for the lead

  • Repair and case systems

    • Service tickets
    • RMA
    • Repairer access requests

  • 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 file

Six layers, tightening inward. Whatever reaches the centre is described in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modePull the agent back to file assembly only when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt and window-rule changes; the Art. 7 platform is due 31 July 2027, foreseen January 2028.Track
L4TraceabilityRecord the order, the clocks, the window applied and each read of the file.Record
L3Named-person closeHold the file for a named human; it governs closing, and not whether a repair is possible.Gate
L2Policy guardrailsTest the file against the windows and part classes on record; no repair information form is mandated.Restrict
L1Confidence thresholdsRoute a thin file to a counsel read before the product compliance lead sees it.Require review
Model coreFile assembled — the order, the clocks, the window owed and completeness
L1 – L2
Test whether a file may stand
L3
Puts the close in a person's hands
L4 – L5
Keep the order and the window behind it
L6
Holds the order at unpromised when signals degrade

How Nestack evaluates it

Evaluate the whole assembly — not only the file that comes out.

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

Surface — the file an authority reads

  1. E1Final-output evaluationDid the file record the window the rule owed that day?
  2. E2Step-level evaluationDid the agent use the right part class, the right audience and the current clock?
  3. E3Tool evaluationDid it read and write the correct order and the correct model record?
  4. E4Confidence calibrationDo low-confidence files actually attract more counsel corrections?
  5. E5Slice evaluationHow does performance change across specific model lines?
  6. E6Business outcomeHow many files needed a correction before the lead closed them?

Floor — the outcome the manufacturer answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, set where each one first enters the availability file.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • IJ-03

      Superseded window rule

      The window applied is the one that model owed last year.

    Stage gathers

    The order, the model file, the clocks and the part classes

  2. 02 · Reasoning

    2 modes

    • IJ-04

      Window read from a target

      A service level stands in for the window owed.

    • IJ-06

      Audience class dropped

      An end-user part is held to be professional-only.

    Stage proposes

    The window, the class applied and completeness

  3. 03 · Tool / write

    2 modes

    • IJ-02

      Clock started from picking

      The clock runs from the pick, not from receipt.

    • IJ-05

      Gear change missed

      The five-day window is run on into year six.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • IJ-01

      Delivered, window unrecorded

      The file shows a delivery but not what was owed.

    Stage returns

    The file a person closes and an authority reads

  5. 05 · Change / Version

    1 mode

    • IJ-07

      Silent window regression

      A rule change moves the window without moving the file.

    Stage tracks

    Model, prompt, window rules and part classes

  • Sev-1 · an order is promised without a person
  • Sev-2 · a wrong window reaches the file
  • Sev-3 · source degrades, file holds unclosed
See the complete Electronics failure-mode catalogue

Affected slices

One model line absorbs the corrections

A model-level window-adherence figure can read clean while a single model line absorbs most of the corrections. Nestack reports the counsel-correction rate by model, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Models in the final two years6.8%3.7×Review
Foldable and hinged models5.0%2.7×Review
Current-generation handsets3.1%1.7×Watch
Slate tablets in early years1.8%1.0×Normal

Bar: counsel-correction-rate lift vs. early-year slate-tablet baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

What an unreconstructable file costs

A cycle closes when the missed delivery window is a regression case. That suite is what the next order measured is judged against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Counsel-correction rate rises in one model line.

  2. 02

    Diagnose

    The part promised in five days by a clock that had already moved to ten is read back until one cause remains.

  3. 03

    Improve

    Changes go out numbered, and the models that drove them ride along.

  4. 04

    Verify

    One model case still failing keeps the release where it is.

  5. 05

    Learn

    It joins the suite for good, and the window rules move in the same commit.

  6. Learn → Detect

    The return edge. The next order is measured against a suite one case longer.

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, file assembly, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Repair-obligation discovery and boundary definition.Week 1
  2. 02Parts, PLM and service sources.Week 1, Week 2
  3. 03Availability-window and part-class rule mapping.Week 1, Week 2
  4. 04Order ingestion and file normalisation.Week 2
  5. 05Model, part-class and audience binding.Week 2, Week 3
  6. 06Completeness scoring and counsel routing.Week 3
  7. 07Named-person close workflow.Week 3, Week 4
  8. 08Parts and service-system integration.Week 5
  9. 09Window and availability cases.Week 4, Week 5
  10. 10Guardrails and approval controls.Week 4
  11. 11Order-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 model line, one marketProductionProduction parts workflowAdvancedMultiple lines / markets
Introduced at Pilot
File assembly to your rules✓✓✓
Named-person close✓✓✓
Availability-period baseline✓✓✓
Introduced at Production
Reporting by model—✓✓
Close workflow in your systems—✓✓
Approved write-back—✓✓
Parts-system integration—✓✓
Introduced at Advanced
Multi-market repair duties——✓
Cross-model window packs——✓
High order volume——✓
Multi-market obligation 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 model masters and parts data

    Order ingestion and part-class mapping

    Week 1

  2. 02Representative orders and model files

    File baseline, window binding and evidence capture

    Week 2

  3. 03Your window rules and part classes

    Window-rule, part-class and audience mapping and the automation boundary

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Parts, PLM and service-record assessment, then integration setup

    Week 2

  5. 05Orders you would not want promised

    Window cases and failure testing

    Week 4

  6. 06What no availability record may promise

    Completeness scoring, counsel routing, guardrails and close controls

    Week 3

  7. 07A named product compliance lead to close

    Named-person close 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 band is the honest length of its phase in working weeks, which is why one week carries two.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Repair-obligation discovery, window mapping and the automation boundary

W2Source integration and the availability baseline

W3File assembly, window logic and close controls

W4Evaluation suite, window cases and failure-mode testing

W5Parts-system integration, pilot orders and targeted corrections

W6One availability year run under the product compliance lead, then Agent Care handover

Reading the band
A bar covers only the weeks its own work is named in. The doubled fifth week is real work, not padding.
At the end of W6
Once the availability file validates, Agent Care picks up the agent.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Electronics AI agent

Build a repair-obligation agent around a clock that changes gear mid-life.

Germany adopted its Reparaturgesetz on 26 June 2026, and the EUR-Lex implementing-measures list could not be read, so we publish no count of Member States that finished. Next step: send us one model file, and we will measure last month's orders against the window each was owed.

Typical delivery
4–6 weeks
Deployment
Your parts system
Named approver
Measured only
After launch
Agent Care available

Nestack Agents · repair obligation & parts

AGT-EL-18 · Agent Care available after launch

Build a repair-obligation agent

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

About you
Your models

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