Main site

Biotechnology AI agent · Trial disclosure

Clinical-Trial Disclosure AI Agent

Keep the disclosure record current: which regime each trial sits under, the clocks that follow, and what the registry record still lacks — then hold it for the named responsible party, who submits.

Typical delivery
4–6 weeks
Deployment
Your stack
Party submits
Pre-submission
After launch
Agent Care

What this agent does

Keeps the record current, never submits it

In

  1. 01

    A study enrols its first participant, and the registration clock runs from that date, not from site activation.

  2. 02

    A protocol amends, and the registry record is diffed field by field against the version that superseded it.

Reason

  1. 03

    A primary completion date is reached, and the results clock is read from that date, not from study completion.

  2. 04

    A results package is assembled, and the elements the rule enumerates are checked present or named as missing.

  3. 05

    A trial ends in a Member State, and the notification clock runs there, the Article 37 results clock from the last one.

Decide

  1. 06

    A record ages, and the verification and update clocks under 42 CFR 11.64 are counted from each element changed.

  2. 07

    A certification falls due, and the evidence for each trial treated as in scope is gathered into one file to sign.

Out

  1. 08

    A Pre-Notice arrives, and the cure period is tracked from receipt against the actions still open on the record.

  2. 09

    A write action runs only inside the approval boundaries agreed during implementation.

Product statement

The agent assembles and keeps the record current; the named responsible party registers, submits and certifies, and the sponsor answers for what the registry shows.

Example workflow

One trial, record to submission

  1. 1Trial record receivedRegistry export, protocol, statistical analysis plan and trial register
  2. 2Record assembledClocks, required elements, protocol diffs and open items, each with its named source
  3. 3Disclosure pack draftedRegistry record, results elements, open items and confidence
  4. 4Controls appliedClock arithmetic, completeness tests against the rule, protocol diffs and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is submitted at any of them — the agent is assembling a record, and the responsible-party lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the responsible party.

Low confidence

Adds a disclosure-lead read first.

Responsible party

The pack is held with its source records, its open items and the confidence.

Approve · Amend · Send to disclosure review

Approved — released to submit

  1. 6Disclosure register updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedAmendment distance, open-item outcomes, registry acceptance and post-posting corrections

Amendments

Each amendment the responsible party makes is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Submitting a registration or a results record.
    • Deciding whether a trial is an applicable clinical trial.
    • Signing the certification of compliance.
    • Requesting an extension, or appealing a refusal.

  • Automation boundary

    Agent acts unaided

    • ✓Assemble the registry record against the protocol and the analysis plan.
    • ✓Compute each clock from the dates already on the record.
    • ✓Test the results package against the elements the rule enumerates.
    • ✓Name what is missing or inconsistent, and hold the pack for the party.

    Writes run only inside the boundaries agreed at implementation, and never ahead of a submission.

    • Setting the primary completion date, or revising it.
    • Answering a Pre-Notice or a Notice of Noncompliance.
    • Placing or validating a document in a submission.
  • Changes to clock, scope or approval rules.

Example output

One disclosure item, annotated

Everything the agent assembles is attached to the record it was drawn from.

Disclosure record output · single trial

Illustrative example
Trial
Interventional, US site
Disclosure item
Results due one year from the actual primary completion date
Time to deadline
138 days
Source of record
Registry record on file
Confidence
88%
Party of record
Responsible party, named on the record

As received

Taken from the registry record and the protocol on file — nothing on this side is written by the agent.

Source facts used

  • Registry record, dated
  • Protocol, version on file
  • Analysis plan section

Why this is open

The clock runs from the primary completion date — 42 CFR 11.44(a).

What the score decides

Below the configured threshold the item picks up a disclosure-lead read before the party sees it.

Action

ApproveAmendSend to disclosure review

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 trial

From the registry record

03Assembly

Read the record against the rule

The protocol is read for what the registry record must say — not for what should have happened to a subject at a site.

01Approved path

The responsible party certifies

The obligation is personal under 42 CFR 11.4. A named person signs Form FDA 3674; the agent builds the evidence file behind it and signs nothing.

02Human review

A posted record is corrected in public

After a record is posted, an error is corrected on the clock 42 CFR 11.64(b) sets, in a registry anyone can read, under the name of the party who submitted it.

04

Build an evidence trail

The record, the clock it was measured against and the party who certified stay on the registry.

Integrations

Typical integrations

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

  • Public registries

    • ClinicalTrials.gov PRS
    • CTIS
    • EudraCT

  • Clinical data systems

    • Medidata Rave
    • Veeva
    • Oracle Clinical One

  • Study documents

    • Protocol
    • SAP
    • CSR
    • Trial master file

Agent

Trial disclosure record

  • Reads the record
  • Assembles the pack
  • Holds for the party

  • Registers and submissions

    • Trial register
    • CTMS
    • Regulatory publishing

  • 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 registry

Six rings, outermost first. Whatever clears the innermost one is named in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeFall back to record assembly when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, clock-rule and regime-configuration changes.Track
L4TraceabilityRecord the sources, the clocks, the open items and the approval time.Record
L3Party approvalHold the pack for the named responsible party; it governs release, not whether the record is right.Gate
L2Policy guardrailsTest the pack against the clock and completeness rules; a failure returns the pack.Restrict
L1Confidence thresholdsRoute low-confidence items to a disclosure-lead read before the party sees them.Require review
Model corePack assembled — registry record, results elements, open items and confidence
L1 – L2
Test whether the record may stand
L3
Puts the release in the party's hands
L4 – L5
Keep the record and the clock behind it
L6
Holds the record at unposted when signals degrade

How Nestack evaluates it

Evaluate the assembly workflow — not only the record that gets posted.

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

Surface — the record the registry shows

  1. E1Final-output evaluationDid every element trace to the record or document behind it?
  2. E2Step-level evaluationDid the agent use the right dates, clocks and regime configuration?
  3. E3Tool evaluationDid it read and write the correct trial and the correct field?
  4. E4Confidence calibrationDo low-confidence items actually attract more party amendments?
  5. E5Slice evaluationHow does performance change across specific trial types?
  6. E6Business outcomeHow many items were amended, or corrected after the record was posted?

Floor — the outcome the sponsor answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, each set where the record first goes wrong.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • EG-03

      Stale registry export

      The record read predates the amendment it is diffed against.

    Stage gathers

    Registry records, protocols, clocks and certifications

  2. 02 · Reasoning

    2 modes

    • EG-04

      Clock read from the wrong date

      Study completion is used where the rule says primary.

    • EG-06

      Backstop lost behind a review

      The outer limit passes while an approval is awaited.

    Stage proposes

    Registry record, results elements and confidence

  3. 03 · Tool / write

    2 modes

    • EG-02

      Cure period never started

      A notice reaches a mailbox nobody now reads.

    • EG-05

      Three registers, three counts

      One portfolio is counted three different ways.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • EG-01

      Missing results element

      An enumerated element is absent from the package.

    Stage returns

    The record the responsible party submits and signs

  5. 05 · Change / Version

    1 mode

    • EG-07

      Silent clock drift

      A model or rule change alters which date starts a clock.

    Stage tracks

    Model, prompt, clock rules and regime config

  • Sev-1 · a wrong record reaches a signature
  • Sev-2 · a wrong date reaches the registry
  • Sev-3 · source degrades, item goes to review
See the complete Biotechnology failure-mode catalogue

Affected slices

A portfolio total can hide one trial type

A portfolio-wide on-time posting rate can read as healthy while a few trial types carry most of the returned items. Nestack reports the return rate by trial type, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Partner-run and licensed trials9.4%3.7×Review
Trials with a delay certification7.6%3.0×Review
Historical CTIS records4.1%1.6×Watch
Single-sponsor US-only trials1.7%0.7×Normal

Bar: return-rate lift vs. single-sponsor baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

Each cycle ends with one more test

A cycle closes when the missed posting is a regression case. That suite is what the next record submitted is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Return rate rises for one trial type.

  2. 02

    Diagnose

    The study that closed in March and posted the following spring is read back until the cause narrows to one.

  3. 03

    Improve

    The change ships with a number, and the trials that forced it travel beside it.

  4. 04

    Verify

    Nothing releases while one touched registry case is still red.

  5. 05

    Learn

    It is kept for good, and the clock rules are amended in the same commit.

  6. Learn → Detect

    The return edge. The suite the next detection runs against is 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, registry workflow, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Disclosure workflow discovery and boundary mapping.Week 1
  2. 02Registry and trial-source assessment.Week 1, Week 2
  3. 03Scope, clock and results-completeness mapping.Week 1, Week 2
  4. 04Record ingestion and reconciliation.Week 2
  5. 05Pack assembly and source binding.Week 2, Week 3
  6. 06Confidence scoring and exception routing.Week 3
  7. 07Responsible-party review workflow.Week 3, Week 4
  8. 08Registry and document integration.Week 5
  9. 09Clock and completeness cases.Week 4, Week 5
  10. 10Guardrails and certification controls.Week 4
  11. 11Registry-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 programme, one registryProductionProduction registry systemsAdvancedMultiple regions / registries
Introduced at Pilot
Disclosure record and reconciliation✓✓✓
Responsible-party approval✓✓✓
Record-completeness baseline✓✓✓
Introduced at Production
Reporting by registry—✓✓
Exception workflow in your systems—✓✓
Approved write-back—✓✓
Registry-platform integration—✓✓
Introduced at Advanced
Multi-regime clock rules——✓
Multi-stage disclosure approvals——✓
High trial count——✓
Multi-regime disclosure controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, trial count, 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 trial list and the regimes they sit under

    Scope, clock and results-completeness mapping

    Week 1

  2. 02Representative prior disclosure packages

    Record ingestion, reconciliation and the assembly baseline

    Week 2

  3. 03Your protocols, analysis plans and study reports

    Disclosure-record mapping and automation-boundary definition

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Registry and document assessment, then integration setup

    Week 2

  5. 05Records you would not want re-read

    Deadline cases and failure-mode testing

    Week 4

  6. 06What no certification may cover

    Confidence scoring, item routing, guardrails and certification controls

    Week 3

  7. 07Named responsible parties to review the pack

    Responsible-party review 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

The bands here follow the weeks each phase truly fills, and that is why the fifth carries two of them.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Disclosure workflow discovery, rule mapping and the automation boundary

W2Registry integration and the reconciliation baseline

W3Assembly workflow, clock logic and review controls

W4Evaluation suite, clock checks and failure-mode testing

W5Document integration, a pilot programme and targeted corrections

W6One posting cycle run under the responsible party, then Agent Care handover

Reading the band
A bar spans the weeks its own work is named in, and no others. The week 5 overlap is real, not padding.
At the end of W6
Validation closes on live trials, and Agent Care picks up the monitoring.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Biotechnology AI agent

Build the disclosure record before the next clock falls due.

Show us your trial list, your clocks and who signs. If a trial in your portfolio was judged out of scope by somebody who has since left, then that judgement is on nobody's file, and the certification signed over it cannot be audited by the person who signed it. FDA wrote to sponsors and researchers on 30 March 2026 about registered trials with no results submitted, and announced it on 13 April 2026.

Typical delivery
4–6 weeks
Deployment
Your registries and CTMS
Party submits
Pre-submission
After launch
Agent Care available

Nestack Agents · Clinical-trial disclosure

AGT-BT-13 · Agent Care available after launch

Build a trial disclosure agent

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

About you
Your trials

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