Back to resources
Research

Regulatory affairs, quality, clinical operations, legal, IT security

What an overnight protocol-amendment review looks like

What does one regulated document workflow look like when the engine cannot file anything?

Morning queue · this walkthrough
1 ready2 exceptions1 held
Last night’s history test · cyan ships
IB silence = protocol errorFailed history
Tie gaps to changed clausesPassed
Require ICF on inclusion changePassed
Clause-gap + ICF mergeMerged
Adaptive controlled-doc boundaryWinner

Counts are from the walkthrough file below — a sample Monday morning, not a promised customer result.

Life sciences is not one workflow. This article is one: a protocol amendment that has to be compared to the last approved protocol, the investigator brochure, the consent form, and the pharmacy manual before regulatory affairs will look at it. Overnight Engine assembles that packet and lists the gaps. It does not decide the science, and it does not file with an IRB, ethics committee, or FDA.

Overnight Engine

On-premAnika Rao · Regulatory affairs lead8:16 AM
1Ready
2Exceptions
1Held

Amendment 3 is assembled against the prior protocol, IB, ICF, and pharmacy manual. Two controlled documents are missing. The packet is ready for RA — not for filing.

Veeva VaulteTMFSharePointTrial master index

HLX-204 A3 · Missing updated ICF

Needs attention

Gaps

HLX-204 A3 · Pharmacy manual 6.3 stale

Review

Gaps

HLX-204 A3 · Inclusion redline ready

Ready

Compare

HLX-204 A3 · IB does not mention age 70

Review

Compare

Helix Therapeutics, Monday 8:16 a.m. Amendment 3 raises inclusion age to 70 and adds a visit-4 blood draw. The ICF is missing from eTMF. Pharmacy manual 6.3 is stale. Anika Rao still owns the review.

Why this article is narrow on purpose

Regulatory packets, quality records, pharmacovigilance, and clinical operations each have different systems, different sign-offs, and different ways a mistake becomes a finding. Treating them as “life sciences AI” is how you get a manifesto instead of a deployment.

Protocol-amendment review is a good first workflow because the corpus is finite, the comparison is direct, and RA already has a completeness checklist. If this loop works, the next article is a different loop — not a broader claim.

What enters the system

HLX-204 Amendment 3 landed in Veeva Vault on Sunday.

  1. 01

    The amendment

    Changed clauses: inclusion age 18–70, visit-4 safety labs, concomitant-med window 14 days.

  2. 02

    The current approved set

    Protocol v2.1, IB v4.2, ICF v2.0, pharmacy manual v5.1, prior IRB correspondence.

  3. 03

    The index

    Helix’s trial master index: which controlled documents must move when inclusion or dosing changes.

What happens overnight

  1. 01

    Intake

    The amendment is read and split into change records, each pointing at a section.

  2. 02

    Compare

    Each change is redlined against the prior protocol and checked for a corresponding IB, ICF, or pharmacy passage.

  3. 03

    Gap list

    Missing or stale controlled documents are listed with the clause that triggered them. Silence in the IB is a note, not an automatic error — Helix history rejected that leap.

  4. 04

    RA packet

    Anika gets a packet she can review. No IRB package and no IND amendment is created.

What regulatory affairs sees in the morning

The age change is cleanly redlined. The visit-4 draw is flagged because IB section 6 does not mention it. The ICF is absent from eTMF. Pharmacy 6.3 still describes the old visit schedule.

Anika can send clinical ops back for the ICF, or she can accept a documented deviation. Those are RA decisions. The engine’s job was to make sure she did not discover the missing ICF on Thursday.

Automation boundaries

Can prepare

  • Redline changed clauses against the prior protocol
  • List required companion documents from the index
  • Assemble an RA review packet with source links

Can execute

  • Write the gap list into Vault as a draft quality record, if Helix enables that write
  • Notify the document owner on an approved template

Always needs approval

  • Scientific or medical judgment on the change
  • IRB / EC / FDA / competent-authority filings
  • Site circulation of the amendment
  • Closing a gap without a person

How a correction becomes a validated improvement

If Anika says an IB silence is expected until the next IB revision, that example stops the engine from calling every silence a protocol error. If she says an inclusion change without an ICF is always a blocker, that rule can win a history test and ship.

Failed hypotheses do not become SOPs. Winning ones change what is on the morning gap list.

What a 30-day pilot measures

One protocol family. Volume will be small; quality of the gap list matters more than throughput.

MetricHow it is measuredPilot target
Time to RA-ready packetAmendment upload → RA can review with gaps listedNext business morning for a standard amendment
Gap precision% of listed gaps RA agrees were realHigh; false gaps train people to ignore the list
Missed companionsRequired docs RA later found that were not listedZero on the pilot protocol family after week two
Unauthorized filingsSubmissions created or sent by the engineZero. Stop-the-pilot metric.

Implementation

Systems
Veeva Vault or equivalent, eTMF, the trial master index, and the SharePoint or QMS home of SOPs.
Data required
Prior amendments, their companion documents, and RA’s actual completeness comments — not just the SOP.
Deployment
Hardware in the GxP-appropriate environment Helix names. Shadow on historical amendments, then supervised packets, then optional Vault writes.
Who operates it
RA owns the review. Clinical ops owns missing documents. Quality owns whether a gap becomes a record. We do not author protocols.

Security and deployment architecture

Protocols, IBs, and consent forms stay inside Helix. Learning is local. If you need a Part 11 or Annex 11 discussion, that belongs in the architecture brief and the quality agreement — not in a generic chatbot clause.

When you want the next article, pick one: deviation / CAPA packets, ICSRs, or TMF completeness. Each gets its own boundaries.

More workflow briefs

Next step

Request a protocol-review assessment

One protocol family, one controlled-document set. Quality, PV, and clinical ops are separate articles when you want them.