← MJM QAAI home

Completed de-identified field case study

EDI cleaning closed with evidence-supported service assurance.

A completed field execution combined practical chemical adaptation, operator interaction, evidence capture, qualified-design measurement traceability, and bounded human approval.

Assurance classClass 2

Service Completion Assurance

Evidence levelLevel 2

Evidence Supported + design traceability

DispositionClosed

Successful provider scope

Customer boundaryClient-held

Executed IOQ and release remain separate

Challenge

Field conditions did not match every draft assumption.

The team needed to interpret a manufacturer manual, execute practical work orders, evaluate commercially available chemicals, adjust calculations, neutralize waste, and keep the operator moving without losing evidence or decision boundaries.

Result

The equipment produced very low-conductivity water after cleaning.

Final neutralized pH was 6.2 after additional recirculation. Final DI rinse water was clear and flowed well through all ports. During post-cleaning operation, the qualified-design process HMI indicated 0.690 µS/cm and its Mettler Toledo M300 conductivity transmitter indicated 0.687 µS/cm. Equipment records and the draft IOQ trace both values to the specified 0–10 µS/cm ROEDI conductivity loop. The readings met the project lead-confirmed site USP <645> conductivity basis; executed qualification and final release remain client-controlled.

Graphical interpretation

Simple field capture, scalable quality assurance.

Class describes what MJM QAAI concludes. Evidence level describes how strongly the event is supported.

Class 1Capture

What was requested, performed, and reported?

Class 2Service assurance

Was the approach sound and the result supportable?

Class 3Controlled conformance

Were controlled requirements, criteria, records, and deviations satisfied?

1SourceManual, work scope, or customer requirement
2Field captureVoice, photos, readings, tickets, and observations
3AI reviewExtract facts, check calculations, identify gaps and substitutions
4Human decisionApprove, hold, condition, or escalate within assigned authority
5CloseoutCertificate, full report, risks, and customer handoff

What the live test taught us

Good control should reduce operator burden.

  • Enter each fact once and reuse it globally.
  • Let technicians submit product and instrument photographs.
  • Make calculations adjustable with visible assumptions and approval history.
  • Keep routine service lean while activating higher controls when risk requires them.

Future development

A separate Phase 2 can test mobile-first execution.

  1. Refine the cleaning procedure and remove redundant entry.
  2. Build the shared mobile/desktop work-item interface.
  3. Test voice, photo, calculation, evidence, and approval workflows.
  4. Re-execute an EDI cleaning and evaluate whether Class 2 occurs naturally.
  5. Pursue Class 3 only where controlled conformance is actually required.

Peer review lens

What this case demonstrates—and what it does not.

Demonstrated

  • Field facts can be captured once and reused across records.
  • AI can support calculations, source interpretation, gap detection, and closeout assembly.
  • A bounded assurance decision can distinguish provider completion from customer qualification.
  • The live test produced measurable improvement and actionable workflow lessons.
  • Two displayed post-cleaning conductivity values from the specified ROEDI measurement loop agreed within 0.003 µS/cm at approximately 0.69 µS/cm.

Not claimed

  • This service closeout does not itself perform or approve cleaning validation or equipment qualification.
  • No customer Quality release or regulatory approval is implied.
  • The assurance model remains an MJM QAAI trial, not an accredited certification scheme.
  • The unsigned certificate preview is retained as a development artifact, not a public credential.
  • The local IOQ is a draft; the executed qualification, current calibration status, alarm disposition, and formal customer release remain client-controlled.

Evidence custody

An audit trail without becoming the client data archive.

MJM QAAI preserves evidence identity, relevant locations, extracted facts, review logic, conclusions, approvals, and hashes where appropriate. Executed qualification packages, calibration records, and controlled raw data can remain in the client system of record.

  1. ReferenceIdentify the controlled source, revision, and relevant section.
  2. ReviewRecord facts, calculations, criteria, gaps, and human decisions.
  3. ManifestRetain a compact evidence trail and permitted de-identified examples.
  4. MinimizeReturn or delete temporary client files under the agreed retention rule.
Disclosure boundary

This case study documents an MJM QAAI trial assurance disposition. It is not a cleaning validation, equipment qualification, regulatory approval, or customer Quality release. Client and site identifiers are intentionally withheld.