Business Central + medtech AI
AI for Dynamics 365 Business Central in Medical Device Manufacturing
Short answer
AI for Business Central in medical device manufacturing helps quality and operations teams retrieve lot and serial genealogy, assemble device history record evidence and prepare audit packages from ERP data and controlled documents, using a private model with every answer tied to its source records. It supports QMSR and ISO 13485 workflows but does not replace a validated QMS or regulatory judgment.
- ERP
- Dynamics 365 Business Central
- Industries
- Medical Devices, Life Sciences, Contract Manufacturing
- Written for
- Quality Manager
Small and mid-size device makers often run Business Central for inventory, production, purchasing and finance while the quality system lives elsewhere: an eQMS such as a dedicated document and CAPA tool, or controlled documents on a share. The ERP holds lots, serial numbers, production orders and supplier data; the QMS holds procedures, complaints, nonconformances and approvals. The two meet at audit time.
That seam is where time disappears. A device history record request, a complaint investigation or a supplier audit needs records from both systems, and the person assembling them has to know which Business Central entries correspond to which QMS record. Native Business Central has no device history record or design history file concept; it provides tracking, change logging and posted ledgers that support one if configured consistently.
Regulation is also shifting. FDA's Quality Management System Regulation, which amends 21 CFR Part 820 and incorporates ISO 13485:2016 by reference, took effect in February 2026, and EU MDR keeps its own demands for traceability and post-market surveillance. Computer systems that touch quality records must be validated for intended use, which makes any AI layer something to scope carefully rather than sprinkle in.
This page lays out a conservative approach: read-only AI over ERP and controlled documents, hosted privately, producing drafts and retrieval for qualified staff. It describes what is helpful, the objects involved, and what remains your quality system's job.
What usually gets in the way
The problems we hear most from quality manager teams running Dynamics 365 Business Central.
Genealogy spans systems
Serial and lot history is in Business Central, while the nonconformance or complaint that triggered the question is in the QMS. Linking them is manual, which slows complaint and field-action investigations.
Device history record assembly is tedious
Collecting production order data, component lots, test results, labeling and release approvals for each unit or batch takes hours when pieces live in different places.
Supplier control evidence is scattered
Approved supplier status, certificates, audit reports and incoming inspection results are spread across vendor records, attachments and files, making periodic reviews slow.
Change control touches many records
An engineering change affects BOM versions, routing versions, open orders and stock of old revisions. Verifying the effect in the ERP is a separate exercise from the change record.
Validation burden discourages tooling
Every new tool touching quality data needs documented intended use and risk-based assurance, so teams avoid helpful tools or adopt them informally, which is worse.
Where AI earns its place in Dynamics 365 Business Central
Each use case names the ERP objects it reads or writes, so your ERP team can judge the integration effort before anyone commits budget.
Lot and serial genealogy retrieval
Given a finished serial or lot, the assistant lists component lots, supplier receipts, production order and test or inspection entries, with each source linked, as a draft for the quality engineer to verify.
Touches: Item Ledger Entry, Serial No. Information, Lot No. Information, Item Application Entry, Production Order
Outcome: Faster first-pass genealogy for complaint and field-action investigations; the engineer reviews and signs.
DHR evidence compilation
For a batch, the agent assembles production order data, consumption lots, operator and work center entries, labeling references and release dates into a checklist against your DHR template and flags missing items.
Touches: Production Order, Prod. Order Component, Capacity Ledger Entry, Item Tracking Line, Document Attachment
Outcome: Shorter review cycles and earlier detection of missing entries before release.
Approved supplier review
Quality asks which suppliers of critical components lack a current audit, certificate or approved status and receives a sourced list.
Touches: Vendor, Item Vendor, Purchase Receipt Line, Document Attachment, Vendor custom fields
Outcome: Periodic supplier reviews prepared from data, not memory.
Change impact check
Before an engineering change is implemented, the assistant lists affected BOM and routing versions, open production orders, stock of superseded revisions and purchase lines in flight.
Touches: Production BOM Version, Routing Version, Prod. Order Line, Item Ledger Entry, Purchase Line
Outcome: Fewer mixed-revision builds and a documented impact summary attached to the change record.
Audit preparation Q&A
Staff ask audit-style questions such as how a given purchase is controlled; the assistant retrieves the relevant procedure sections and sample ERP records with citations.
Touches: Controlled documents index, Change Log Entry, Purchase Header, Item Tracking Line
Outcome: Auditor questions answered with references faster, with staff verifying the answer.
Expiry and sterility date monitoring
A daily read of lots with expiration dates by location and open demand, flagging stock that cannot ship within remaining shelf life requirements.
Touches: Lot No. Information, Item Ledger Entry, Bin Content, Sales Line
Outcome: Less expired stock discovered late.
UDI and labeling data checks
The agent compares item GTIN and UDI attributes held in item or custom fields against labeling records and flags inconsistencies for review.
Touches: Item, Item Reference, custom UDI fields, Item Ledger Entry
Outcome: Inconsistencies found before shipment rather than during an audit.
Reference architecture
A private, read-only layer retrieves from Business Central and from your controlled-document store, produces cited drafts and never writes to quality records. The design is built so it can be described in an intended-use statement and assured proportionately to risk.
- 1
ERP read interface
Scoped API pages over ledger, lot, serial, production and vendor data, with Change Log Entry exposed where audit trail review matters.
- 2
Controlled document index
Procedures, work instructions and specifications indexed with document ID, revision and status, so only effective revisions are cited.
- 3
Citation and traceability
Each statement carries a pointer to the ledger entry or document revision it came from; unsupported answers are declined.
- 4
Model hosting
Open-weight models hosted privately, with version pinning so behavior is stable between validations.
- 5
Assurance and logging
Intended-use statement, risk assessment, test set and prompt logs retained to support a risk-based assurance approach.
Integration notes for your ERP team
- Expose Serial No. Information and Lot No. Information along with Item Application Entry; genealogy from lot numbers alone is incomplete.
- Turn on and expose the Change Log for tables where audit trail matters, and set a retention policy since it grows quickly.
- Document the intended use of the tool: read-only retrieval and drafting, no record creation or approval.
- Pin model version and prompt templates; re-run the test set before any change, and keep results as assurance evidence.
- Integrate with the eQMS through its API or export only for read access, matching records by item, lot or NCR number.
- Plan for Business Central release waves: test custom API pages in sandbox preview before each wave.
Deployment options
Dedicated private environment
Device makers with strict IP and design confidentiality
Model and index run in a locked-down subscription or on-site server; Business Central is accessed through scoped service credentials.
On-prem with Business Central On-Premises
Contract manufacturers with customer-imposed data limits
All components inside the plant network with no external model calls; updates are scheduled and documented.
Staged pilot on a sandbox copy
Teams that must complete risk assessment before touching production
Begin on a sandbox environment with masked data, finalize intended use and tests, then switch the read connection to production.
Compliance and data control
How the architecture supports your obligations. Certification and accountability stay with your organisation; the design keeps the evidence straightforward.
FDA QMSR (21 CFR Part 820) and ISO 13485:2016
The assistant supports retrieval and compilation of records; procedures for document and record control remain the manufacturer's responsibility.
21 CFR Part 11 electronic records and signatures
The AI does not sign or approve; Business Central's change log and your QMS handle audit trails, and AI outputs are treated as drafts.
FDA Computer Software Assurance guidance
Risk-based assurance for the tool's intended use, with documented test cases; version pinning of models.
EU MDR and UDI requirements
Traceability and UDI consistency checks assist post-market surveillance data gathering; regulatory strategy stays with your team.
Design confidentiality
Private hosting keeps designs, suppliers and process data inside your boundary.
Where Netray fits
ERPray
Cited genealogy, supplier and expiry questions over Business Central with permissions respected.
DataRay
Searching controlled documents and audit files on local shares together with ERP records.
Custom build
DHR compilation workflows and integrations with your eQMS under a documented intended-use statement.
How an engagement runs
Phase 1 . 1 to 2 weeks
Scope and intended use
- -Intended-use statement and risk assessment draft
- -Records and systems map (ERP and QMS)
- -Priority retrieval scenarios
Phase 2 . 3 to 4 weeks
Sandbox build
- -Read-only endpoints and document index
- -Test cases with expected answers
- -Citation behavior review
Phase 3 . 3 to 4 weeks
Assurance and pilot
- -Executed test evidence
- -Pilot with quality engineers
- -Access and logging review
Phase 4 . 2 to 3 weeks
Controlled release
- -Production read connection
- -Change procedure for model and templates
- -Training and runbook
Questions to ask any vendor, including us
A short list that separates real Dynamics 365 Business Central AI work from a chatbot demo.
- How does the vendor support our computer software assurance approach?
- Is the tool strictly read-only, and how is that enforced?
- Does every statement include a citation to a record or controlled document revision?
- How is the model version controlled and re-tested after changes?
- Where do designs, supplier data and complaint details get processed?
- How does it handle Business Central release waves?
- What happens when the answer is not supported by records?
Frequently asked questions
Can an AI tool be used in a regulated medical device environment?
Yes, with discipline. Define the intended use, assess the risk, test against known cases, control versions and keep humans accountable for decisions. A read-only retrieval and drafting tool carries less risk than one that creates or approves records. Your quality team and regulatory advisor decide what assurance is sufficient.
Does Business Central include a device history record?
Not as a named feature. It provides production orders, tracking, ledgers and a change log that can support a DHR if processes and configuration are consistent. Many manufacturers combine it with an eQMS or controlled documents. The AI helps assemble evidence across them; it does not create the record structure.
What changed with QMSR?
FDA's Quality Management System Regulation amends 21 CFR Part 820 and incorporates ISO 13485:2016 by reference, effective February 2026. For ERP users, the practical effect is alignment with ISO 13485 terms and expectations. Confirm specifics with your regulatory advisor; this page does not provide regulatory advice.
Will it write CAPAs or approve changes?
No. It can retrieve related records and draft summaries that a qualified person reviews. Creating or approving CAPAs, nonconformances and changes should remain in your QMS under your procedures and electronic signature controls. Keeping the AI out of the approval path simplifies assurance.
How do you prevent made-up answers?
By requiring citations to specific ledger entries or document revisions and declining to answer when none support a statement. We also maintain a test set of known questions run after any change. No approach eliminates errors, so staff verify before relying on an answer in a record.
Can we run it fully offline?
Yes. With open-weight models on local GPUs and Business Central On-Premises, the system can run in a network with no internet access. For SaaS tenants, only the scoped API calls reach the tenant, and model processing stays on your infrastructure.
How long does a validated pilot take?
Typically eight to twelve weeks, driven more by intended-use documentation, test design and review cycles than by technical build. Teams with an existing assurance procedure move faster. We recommend starting in a sandbox with masked data while paperwork is completed.
Related guides
AI for Business Central manufacturing operations, beyond Copilot
AI for Business Central manufacturing beyond Copilot: grounded answers on production orders, routings, and capacity via APIs v2.0, on-prem capable.
Business Central + private AIAI for Dynamics 365 Business Central, on-premises or SaaS, beyond Copilot
AI for Business Central beyond Copilot: grounded question answering over BC's data via OData and APIs, for on-premises and SaaS tenants alike.
Faster paperwork, same standardAI for NCR, CAPA, and 8D Drafting in AS9100 and ISO 9001 Quality Systems
AI drafts NCR, CAPA, and 8D reports from ERP quality data for engineer review, cutting drafting time under AS9100 and ISO 9001 without changing sign-off.
MasterControl + private AIAI for MasterControl: Controlled Documents and Quality Events Joined to Your ERP
Private AI for MasterControl QMS: document control, training, CAPA and batch records joined to ERP lots and suppliers, with human sign-off and audit trails.
ETQ Reliance + private AIAI for ETQ Reliance: Joining Your Quality Records to Your ERP Data
Add grounded AI to ETQ Reliance for CAPA, complaints, audits and document control, joined to ERP lots, suppliers and work orders, with private deployment options.
GDPR + AI on ERP dataBuilding GDPR-Compliant AI on Top of Your ERP
Design AI on ERP data that satisfies GDPR: lawful basis, DPIA, data minimisation, and an on-prem architecture that avoids the Schrems II transfer problem.
Plan it with numbers
Cost of Quality (COQ) Calculator
Add up prevention, appraisal, internal failure, and external failure costs to see your total COQ, its share of revenue, and your gap to world-class performance.
Free ToolERP AI Maturity Assessment
Benchmark how deeply AI and automation are embedded in your ERP operations, from data foundations to autonomous agents, across four maturity levels.
Free ToolAir-Gapped AI Readiness Assessment
A 10-question assessment that scores how prepared your organization is to deploy and operate LLMs inside an air-gapped or classified enclave.
GuideERP Audit Trail Configuration Guide
Configure comprehensive ERP audit trails covering transaction logging, change tracking, data retention policies, and audit-ready reporting for SOX and ISO compliance.
GuideAn ERP Data Quality Framework
An ERP data quality framework for manufacturers: profiling item and BOM data, six quality dimensions, automated rules, scorecards, and remediation workflows.
GuideFrom AI Pilot to Production: The Playbook
Move an AI pilot to production: environment promotion, monitoring, rollback plans, and the ERP integration hardening manufacturers need before go-live.
Talk it through with an engineer who knows Dynamics 365 Business Central
Bring one real question your team cannot answer from the ERP today. We will map the data path, the model, and where it runs, and tell you honestly if AI is the wrong tool for it.