Mattec MES + private AI
AI for Epicor Mattec MES: Press Data, Scrap and Mold History in Plain English
Short answer
AI for Epicor Mattec MES connects a private language model to the MES database so plant managers and process engineers can ask about press downtime, scrap by reason, cycle time drift and mold performance, with the query shown. Joined to ERP orders and costs, it runs on-prem next to the MES, which is often hosted on a plant SQL Server.
- ERP
- Epicor Mattec MES
- Industries
- Plastics, Injection Molding, Automotive, Medical Devices
- Written for
- Plant Manager
Mattec MES, now part of the Epicor portfolio, is widely used by injection molders and other process-oriented plants for press monitoring, downtime and scrap capture, production tracking, and mold and tool management. The data model is built around machines or presses, jobs, cycles, downtime events, scrap events and tooling.
Plastics plants run on thin margins and long production runs, so small changes matter. A slow cycle on a high-volume mold, a recurring flash defect, or an unrecorded changeover compounds over thousands of shots. The monitoring data to detect these is already in the MES, but few people have the time to mine it.
The built-in reports are good for known questions: efficiency by press, scrap by reason, downtime by code. The questions that cost money are different: why does this mold run slower on press 7 than press 3, which resin lots correlate with scrap, and what does the ERP say we should have made last week against what the MES recorded.
This page covers how a private AI layer can sit next to Mattec MES and the ERP, which data it needs, how to keep it read-only, and where the standard Mattec reports and Epicor tools remain the better choice.
What usually gets in the way
The problems we hear most from plant manager teams running Epicor Mattec MES.
Scrap reasons are recorded but rarely analysed together
Scrap codes, press, mold, material lot and shift sit in different tables and reports. Linking a defect pattern to a resin lot or a mold revision requires manual joining that happens only after a customer complaint.
Mold and tool knowledge is tribal
Which molds need a specific press, which cavities are blocked, and which maintenance was done last are known by a few setters and toolroom staff and are not easy to retrieve quickly.
Cycle time drift goes unnoticed
A cycle that creeps up by a fraction of a second across a long run is invisible on a daily summary but material over a month; detecting it needs trend analysis per mold and press.
MES and ERP disagree on counts
Good parts, scrap and material consumption in the MES differ from ERP production reporting, particularly when operators correct counts late.
Regulated customers expect traceability
Medical and automotive customers expect lot and process records on demand, which stretches a small quality team during a complaint.
Where AI earns its place in Epicor Mattec MES
Each use case names the ERP objects it reads or writes, so your ERP team can judge the integration effort before anyone commits budget.
Scrap analysis by mold, press and lot
Ask which scrap reasons rise on a given mold or after a resin lot change and get a breakdown with the data behind it.
Touches: Scrap events and reasons, jobs, molds, presses; ERP material lot receipts
Outcome: Faster identification of material or tool related defects, shortening the time between first occurrence and corrective action.
Cycle time drift detection
Compare current cycle time with the standard and the historical median per mold and press, and flag sustained drift.
Touches: Cycle records, standard cycle on job or part, press IDs
Outcome: Earlier notice of issues such as cooling or hydraulic problems; flagged items need a technician to confirm.
Downtime Pareto and changeover review
Rank downtime by reason and compute changeover duration by mold family, highlighting the outliers.
Touches: Downtime events and codes, job start and end times, mold change records
Outcome: A targeted SMED focus list instead of a general call to reduce changeovers.
MES to ERP production reconciliation
Compare good quantity, scrap and material usage in the MES against ERP production postings and list the jobs that disagree.
Touches: MES job production totals, ERP job or production order receipts and issues
Outcome: Month-end variance investigation becomes a short exception list.
Mold maintenance history lookup
Retrieve shots since last maintenance, repair history and known cavity issues for a mold before it is scheduled.
Touches: Mold records, shot counters, maintenance notes, ERP asset or tooling records
Outcome: Setters and schedulers avoid starting a job on a mold that is overdue for service.
Customer complaint traceability pack
Given a part and date range, assemble jobs, presses, molds, material lots and operators involved, for a quality engineer to review.
Touches: Job history, lot records, process data, shipment lines
Outcome: Complaint response preparation shrinks from a day of gathering to a review of one assembled pack.
Shift and plant summary
Draft a daily summary of efficiency, scrap hot spots and jobs behind schedule for the production meeting.
Touches: Efficiency metrics, open jobs, due dates
Outcome: Meeting preparation time falls, with the summary reviewed before circulation.
Reference architecture
The AI layer reads a replica or reporting copy of the MES database, aligns it with ERP jobs and lots through the ERP's APIs, and uses a locally hosted model that queries curated views and cites its sources.
- 1
Data access
Read-only access to a copy of the MES SQL database, scheduled so analytical queries do not affect press monitoring.
- 2
Semantic layer
Business-term views for press, mold, job, cycle, downtime and scrap, with metric definitions such as efficiency or good-part yield fixed once.
- 3
ERP join
Mapping of MES job numbers and part numbers to Epicor or other ERP jobs, items and lots, using ERP APIs for reads.
- 4
Model serving
An open-weight model hosted on a plant or data center GPU, using tool calls to query views instead of recalling facts.
- 5
Controls
Query display, role scoping by plant, logging, and human approval before any change to ERP or MES data.
Integration notes for your ERP team
- Confirm the Mattec version and database layout with Epicor or your integrator before designing queries, as table structures differ between releases.
- Use a SQL Server replica or reporting copy for analysis and a read-only login with access limited to named views.
- Normalise press and mold identifiers across the MES, ERP asset records and maintenance notes.
- Capture resin lot data at the job level if it is not already recorded, as lot-to-scrap correlation depends on it.
- Use Epicor or other ERP APIs for job, item and lot reads instead of querying the ERP database directly where possible.
- Treat standard cycle times as data that needs periodic review, since drift detection is only as good as the baseline.
Deployment options
Plant on-prem
Molders with the MES hosted locally
GPU server beside the MES SQL Server on the plant network, with no external model calls.
Central private cloud
Multi-plant molders
Replicated MES data from each plant lands in one analysis environment, scoped per plant.
Air-gapped
Defense or export-controlled programs
Isolated analysis environment fed by approved transfers.
Compliance and data control
How the architecture supports your obligations. Certification and accountability stay with your organisation; the design keeps the evidence straightforward.
IATF 16949
Process and scrap evidence becomes retrievable for audits and corrective actions; the quality system remains the record.
FDA 21 CFR 820 / QMSR and ISO 13485
For medical molders the AI is a read-only retrieval aid. Validated processes continue to govern device history records.
ITAR / EAR
Where part data is controlled, keeping model and index inside the customer's boundary lets their export program decide access.
Customer quality flow-downs
Answers cite source records, which supports complaint and PPAP evidence requests.
Where Netray fits
How an engagement runs
Phase 1 . 1 to 2 weeks
Discovery
- -Question inventory from production and quality
- -MES version and schema review
- -ERP integration points
Phase 2 . 2 to 4 weeks
Semantic model
- -Curated views for presses, molds and scrap
- -Metric definitions agreed
- -Read-only access
Phase 3 . 3 to 5 weeks
Pilot
- -Q&A for a set of presses
- -Verified answer test set
- -Operator and engineer feedback
Phase 4 . 3 to 5 weeks
Rollout
- -Remaining presses and plants
- -Permissions model
- -Monitoring and handover
Questions to ask any vendor, including us
A short list that separates real Epicor Mattec MES AI work from a chatbot demo.
- Which Mattec version are we on and what is the database layout?
- Can analytics run on a replica without affecting press monitoring?
- Is resin lot recorded at the job level today?
- How will job numbers be matched between MES and ERP?
- Can every answer show its query and underlying rows?
- What does the standard Mattec reporting already do well?
- How will complaint packs be reviewed before they go to a customer?
Frequently asked questions
What is Epicor Mattec MES used for?
It is a manufacturing execution system commonly used by injection molders and similar plants to monitor presses, record downtime and scrap, track production against jobs, and manage molds and tooling. It provides real-time visibility on the floor and feeds production data back to the ERP.
Does Epicor offer AI for Mattec?
Epicor has been adding AI features across its portfolio, so check the current roadmap and licensing for your version. A separate private layer is worth considering if you need cross-system questions, unusual analysis, or want the model and data to remain entirely within your own infrastructure.
Can the AI find the cause of scrap?
It can surface correlations, such as a rise in a scrap reason after a resin lot change or on one mold-press pairing, with the data shown. Confirming a root cause still requires process engineers and sometimes trials. We present correlations as leads, not conclusions.
Does it only work with Epicor ERP?
No. Mattec is often paired with Epicor Kinetic or Prophet 21, but the AI layer joins to any ERP with a usable API or reporting interface. The effort is in matching jobs, items and lots, which is similar across systems.
Will it slow down the MES?
It should not. We read from a replica or scheduled copy of the database, not the live monitoring instance, and use a read-only account limited to specific views. We recommend testing query load during off-peak hours before rollout.
Can the AI write to Mattec or the ERP?
Not by default. The standard design is read-only. If you want it to propose changes such as updated standard cycle times, a person approves them and the change is made through supported interfaces with an audit trail.
How large a GPU do we need?
For a plant team asking questions during the day, a single modest GPU server and a mid-sized open-weight model are often enough. We size from concurrent users and document volume, and some sites begin on shared infrastructure.
Related guides
AI for Epicor Kinetic, Beyond What Prism Covers
Add AI to Epicor Kinetic beyond Prism: private LLM over BAQs, BPM data, and REST v2, on-prem or private cloud, with honest guidance on when Prism already covers you.
Prophet 21 + private AIAI for Epicor Prophet 21, Built for Distribution Operations
Prophet 21 AI for distributors: on-prem or private-cloud LLM grounded in P21 order, inventory, and purchasing data, for order follow-up, forecasting, and pricing questions.
Vantage / Vista / E9 + private AIAI for Epicor Vantage, Vista, and E9, and a Real Kinetic Upgrade Path
Still on Epicor Vantage, Vista, or E9? Add a private LLM over that Progress OpenEdge database now, and use it to plan a Kinetic upgrade with real data, not guesswork.
RAG + SQL + permissionsA Private LLM Grounded on Your ERP Data
How a private LLM answers questions on your ERP data: RAG plus text-to-SQL, role-based permissions inherited from the ERP, and where each fits.
Agents + approval gatesAI Agents for ERP, Running On-Prem
A practical guide to on-prem AI agents for ERP: what they can safely automate, where human approval belongs, and how to design the guardrails.
Opcenter Execution + private AIAI for Siemens Opcenter Execution MES, Running Inside Your Own Network
Add private, grounded AI to Siemens Opcenter Execution MES: ask questions over lots, work orders, nonconformances and genealogy without leaving your network.
Plan it with numbers
OEE Calculator (Overall Equipment Effectiveness)
Break your Overall Equipment Effectiveness into availability, performance, and quality, and see how far you sit from the 85% world-class benchmark.
Free ToolManufacturing Downtime Cost Calculator
Combine lost contribution margin, idle labor, and absorbed overhead into a defensible monthly and annual downtime cost - with recovery effects modeled.
Free ToolShop Floor Digitization Scorecard
A 10-question scorecard measuring how much of your shop floor still runs on paper - and which digitization gaps are costing you the most.
GuideSmart Factory ERP Integration: Industry 4.0 Connectivity Guide
Connect your smart factory systems to ERP with this Industry 4.0 integration guide covering MES, SCADA, IoT platforms, and real-time production data pipelines.
GuideAI Predictive Maintenance with ERP Integration
Integrate AI predictive maintenance with your ERP system. Reduce unplanned downtime by 45% using sensor data, ML models, and automated work order generation.
GuideOn-Prem LLM Deployment Architecture: Reference Guide
Reference architecture for on-prem LLM deployment: inference servers, GPU sizing, RAG pipelines, and security zones for regulated manufacturers.
Talk it through with an engineer who knows Epicor Mattec MES
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.