Accessibility Conformance & Certification AI Agent

What this agent does

Assembles the evidence, not the certification

A title is delivered, and its caption, audio-description and signing tracks are recorded against the catalogue.

A platform is added, and each build is measured on its own — Ofcom's draft Code proposes no averaging across them.

Reason

A caption sample is measured against the §79.1(j)(2) standards, which the 2026 deletion rule left untouched.

A quarter closes, and audio-description hours are counted per §79.3 window rather than annualised and divided.

An election is evidenced, because quality compliance and §79.1(k) Best Practices rest on different records.

Decide

An EU access layer is documented under the European Accessibility Act, which has applied since 28 June 2025.

A draft is routed to the named officer of the programmer, who elects, signs and makes the certification.

Out

A snapshot is kept of what the evidence showed on the day, so the figures reproduce when someone asks.

A write is made only inside the approval boundaries agreed during implementation, never ahead of approval.

→ Product statement

The agent assembles the evidence; a named officer of the programmer certifies, and a distributor or a regulator relies on what was signed.

Example workflow

One reporting year, catalogue to signature

AgentHuman

1Catalogue assembledTitle records, access-service tracks, platform manifests and delivery receipts

2Evidence gatheredCaption, audio-description and signing coverage per platform, with the monitoring record

3Conformance measuredCoverage, measured quality, exemption categories and confidence

4Controls appliedStandard checks, per-platform reconciliation, EU access-layer checks and confidence threshold

No human action required

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

5DecisionBranches at the confidence threshold

High confidence

Goes to the certifying officer to sign.

Low confidence

Adds an accessibility-lead read first.

Officer certification

The draft is held with its evidence, its unsupported lines and the confidence.

Approve · Amend · Send to accessibility lead

Certified — signed and made available▼

6Conformance systems updatedOnly where write access and approval policy allow it

7Outcome evaluatedEvidence gaps, per-platform variance, complaint outcomes and corrections after certification

Amendments

Every amendment the officer makes is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items

Signing or making the certification.

Electing quality compliance or Best Practices.

Asserting a disproportionate burden under the EAA.

Filing an annual report to a regulator.

Automation boundaryAgent acts unaided

✓Maintain the conformance evidence across the catalogue.

✓Measure captions against the codified quality standards for the named owner.

✓Count hours and coverage per quarter and per platform.

✓Mark what the evidence does not support and hold the draft for the officer.

Any write happens inside the boundaries agreed at implementation, never ahead of the signature.

Warranting conformance to a distributor.

Deciding that a claimed exemption holds.

Judging an access feature good enough to count.

Changes to conformance, evidence or approval rules.

Example output

One conformance line, annotated

Everything the agent assembles is attached to the evidence it was measured from.

Conformance output · single titleIllustrative example

Service

Measured line

Platform

Evidence of record

Confidence

Position

On-demand catalogue

Subtitles present on the delivered build, audio description absent

Connected-TV app

Delivery receipt on file

92%

Not supported for certification

As receivedTaken from the catalogue record and the platform manifest — nothing on this side is written by the agent.

Conformance evidence used Platform manifest Monitoring record Vendor delivery receipt

Why this positionThe description track is absent on this build, whatever the master carries.

ActionApproveAmendSend to accessibility lead

What the score decidesBelow the configured threshold the draft picks up an accessibility-lead read before it.

Value

Where AI adds value

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

Where the value landsValue 01 – 04

Every titleFrom the catalogue record

03Measuring

Measure against the standard

Read the delivered tracks, the platform manifests and the monitoring records kept on the §79.1(c)(3) horizon.

01Approved path

The rule is written but not live

The §79.1(m) filing waits on a Federal Register date; the live certification runs to distributors under §79.1(k).

02Human review

Know today what you cannot defend

No EAA action has reached a media provider yet; the unsupported lines are marked before anyone asks for them.

04Build an evidence trail

The title, the standard it was measured against and the officer who certified 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.

Catalogue and metadataMAM · DAM libraries
Access-service metadata

Delivery and platformsCTV apps · set-top builds
Partner and device platforms

Access-service vendorsCaptioning · description
Signing and QC vendors

Agent

Accessibility conformance

Reads the catalogue
Measures conformance
Holds for certification

Compliance recordsMonitoring and complaint logs
Regulatory correspondence

Observability & evaluationOpenTelemetry · 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 certification

Six controls, each enclosing the next. What gets past every one is named in the map below this.

L6 · Outermost — last line of defenceInward → L1 · closest to the model

L6Rollback / safe modeStep back to conformance reporting when evaluation or production signals degrade.Roll back

L5Version monitoringTrack model, prompt, conformance-rule and standard-version changes.Track

L4TraceabilityRecord the title, the evidence, the measurement and the certification time.Record

L3Officer sign-offHold the draft for the named officer; it governs issue, not whether a signed claim is true.Gate

L2Policy guardrailsTest each claim against the evidence held; a claim without evidence returns unsupported.Restrict

L1Confidence thresholdsRoute low-confidence measurements to an accessibility-lead read first.Require review

Model coreConformance measured — coverage, measured quality, exemption categories and confidence

L1 – L2Test whether a claim may stand

L3Puts the signature in an officer's hands

L4 – L5Keep the title and the standard behind it

L6Steps back to conformance reporting when signals degrade

How Nestack evaluates it

Evaluate the conformance workflow — not only the finished certification.

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

Surface — the record the regulator reads

Depth of coverage ▼

E1Final-output evaluationDid each claim in the draft match the evidence held?

E2Step-level evaluationDid the agent use the right catalogue, platform build and standard?

E3Tool evaluationDid it read the correct title and the correct platform manifest?

E4Confidence calibrationDo low-confidence measurements actually attract more amendments?

E5Slice evaluationHow does performance change across specific catalogue cohorts?

E6Business outcomeHow many claims needed an amendment or a correction after certification?

Floor — the outcome the programmer answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, placed at the stage each one originates.

Agent lifecycleDirection of processing →

01 · Retrieval1 mode

UQ-03

Catalogue read as ingested

Coverage taken from the master, not the platform build.

Stage gathersTitle records, platform manifests and delivery logs

02 · Reasoning2 modes

UQ-04

Election unsupported

Best Practices are certified with no process record.

UQ-06

Blended platform figure

Per-platform results are averaged into one number.

Stage proposesCoverage per platform, quality and confidence

03 · Tool / write2 modes

UQ-02

Assembly read as certification

A draft stands where a signature should be.

UQ-05

Wrong clock tracked

An August 2026 display date is read as the filing.

Stage writesOnly where write access and approval policy allow it

04 · Output1 mode

UQ-01

Claim without evidence

A certification asserts what the record cannot show.

Stage returnsThe certification the officer signs and stands behind

05 · Change / Version1 mode

UQ-07

Silent standard drift

A rule change widens what counts as conformant.

Stage tracksModel, prompt, conformance rules and open proposals

Sev-1 · a certification goes out unsupported Sev-2 · a wrong figure reaches the record Sev-3 · source degrades, line routes to review

See the complete Media & Entertainment failure-mode catalogue →

Affected slices

Catalogue conformance can hide one weak platform

A catalogue-wide conformance figure can look acceptable while a small number of platform cohorts carry most of the shortfall. Nestack reports the unevidenced-claim rate by slice, not only in total.

Slice performance — reported separately, not only in aggregateIllustrative example

Slice	Failure rate	Lift	Status
Partner-operated device builds	6.3%	3.5×	Review
Back catalogue and library titles	4.2%	2.3×	Review
Trailers and short-form extras	2.9%	1.6×	Watch
Current-season flagship app	1.6%	0.9×	Normal

Bar: unevidenced-claim-rate lift vs. flagship-app 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 conformance case

A cycle is done when the unevidenced claim has become a case the next release must close. That suite is what the next certification made is measured against.

Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn

01Detect

Unevidenced-claim rate rises in a platform slice.

02Diagnose

The back half of the catalogue nobody had measured is read until one cause is left.

03Improve

The change is numbered on the way out, with the titles that revealed it.

04Verify

A conformance case that has not cleared keeps the release parked.

05Learn

It joins the suite for good, and the conformance rules are edited to match.

Learn → DetectThe return edge. The next detection runs against a suite one conformance 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, conformance workflow, evaluation, integration, then production validation and handover.

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

01Conformance workflow discovery and boundary setting.

02Catalogue and platform source review.

03Caption standard, quota and exemption-rule mapping.

04Catalogue ingestion and evidence mapping.

05Conformance measurement and binding.

06Confidence scoring and gap routing.

07Officer certification workflow.

08Catalogue and delivery integration.

09Conformance and evidence cases.

10Guardrails and certification controls.

11Catalogue-trail instrumentation.

12Deployment, documentation and Agent Care handover.

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 tier PilotOne service, one platform ProductionProduction catalogue systems AdvancedMultiple services / markets

Introduced at Pilot

Measuring to your catalogue and rules✓✓✓

Officer approval✓✓✓

Conformance-evidence baseline✓✓✓

Introduced at Production

Reporting by service—✓✓

Approval workflow in your systems—✓✓

Approved write-back—✓✓

Catalogue and delivery integration—✓✓

Introduced at Advanced

Multi-regime conformance rules——✓

Multi-stage compliance approvals——✓

High catalogue volume——✓

Multi-platform conformance controls——✓

Build price From $5,000 From $8,000 Custom quote

Final build priceConfirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.

Separate from buildBuild 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.

You bringWe build with it

01Your catalogue records and platform builds → Catalogue ingestion and evidence mappingWeek 1

02Representative delivered titles → Conformance baseline and standard measurementWeek 2

03Your access-service policies and vendor terms → Caption standard, quota and exemption-rule mappingWeek 1

04Access to relevant APIs, feeds or exports → Catalogue and delivery assessment, then integration setupWeek 2

05Certifications you would not want relied on → Complaint cases and the evaluation suiteWeek 4

06What no certification may claim → Confidence scoring, gap routing, guardrails and approval controlsWeek 3

07A named officer to make the certification → Officer certification workflow, then pilot and production validationWeeks 5–6

Nothing else is required Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Bands sit where the reporting year puts them, which is why week 5 carries two phases rather than one.

Phase W1W2W3W4W5W6

Discovery W1

Build W2 – W3

Evaluate W4 – W5

Pilot & Launch W5 – W6

Week focus W1Conformance workflow discovery, rule mapping and the automation boundary W2Catalogue and delivery integration and the measurement baseline W3Measurement workflow, confidence logic and approval controls W4Evaluation suite, standard checks and failure-mode testing W5Platform integration, pilot titles and targeted corrections W6One reporting year assembled under the accessibility lead, then handover

Reading the bandA bar spans the weeks its work is named in, and no more. The week 5 overlap is real work, not padding.

At the end of W6Once the year validates, monitoring transfers to Agent Care.

DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Media & Entertainment AI agent

Build an accessibility conformance agent around the certification you sign.

Show us your catalogue, your platform builds and the officer who signs. What you get is the evidence standing ready on the day someone asks, and a list of what you could not defend today.

Discuss this agent →

Nestack Agents · Accessibility conformanceAGT-ME-20 · Agent Care available after launch