How to add a private AI layer to an on-prem CMMS
how to add AI to an on-prem CMMS
Also searched as
- AI for Maintenance Connection on-prem
- MP2 CMMS AI integration
- chat with CMMS work order history locally
- CMMS LLM without cloud
Short answer
Maintenance history is mostly free text and codes, which is what retrieval and a local model handle well. Read the CMMS through its supported API, ODBC views or a reporting copy, index work order text and manuals, and let a local model answer from cited records. Creating or changing work orders goes through the CMMS's own interface with human approval.
Applies to: On-prem CMMS products backed by SQL Server or similar, including Maintenance Connection and MP2 class systems. Many SaaS CMMS products (Fiix among them) expose only an API, with no database access.
Rollout for an AI layer on a CMMS
- 1Confirm what access the vendor supports: a documented web service or REST API, database views for reporting, or neither. Use the supported route first and ask the vendor in writing before reading tables directly.
- 2Identify the objects you need: assets and locations, work orders and tasks, PM schedules, parts and inventory, failure and cause codes, technician notes.
- 3Create a read-only account and a reporting copy of the database if one is permitted. Keep AI queries off the production database.
- 4Export or stream work order history into a retrieval index, with asset, location, craft, failure code and date as metadata, and redact any personal data you do not need.
- 5Add OEM manuals, PM procedures and lubrication standards from file shares to the index, linked to asset classes.
- 6Expose approved tools: find_similar_failures, asset_history, pm_compliance, parts_for_asset, backlog_by_craft, each returning cited record numbers.
- 7Serve an open-weight model locally and limit it to those tools. Log each question, the records returned and the user.
- 8Draft before you write: let the assistant propose a work order description and parts list that a planner reviews and enters through the CMMS.
Data access paths
Packaged on-prem CMMS products usually sit on SQL Server and offer a web service layer, import and export utilities and reporting views. The right route depends on the vendor and version, so confirm which interfaces are supported before building. Direct table reads are the weakest option: schemas change between releases and support may be affected.
Whichever route you use, the data model is similar: assets in a hierarchy, work orders with status and type, tasks, labour and parts transactions, PM schedules driven by calendar or meter, and failure, problem and cause codes. The valuable text sits in descriptions, technician notes and close-out comments.
SaaS CMMS products often have an API but no database access. This article assumes on-prem, where you can also take a reporting copy. If your CMMS is cloud-hosted, the same retrieval design applies but data will have to be pulled through the API into your own store.
-- Illustrative reporting query: recent corrective work orders on one asset
-- (table and column names are generic, map them to your CMMS schema)
SELECT wo.wo_number, wo.opened_date, wo.problem_code, wo.cause_code,
wo.description, wo.completion_notes
FROM reporting.work_orders wo
WHERE wo.asset_id = 'PRESS-014'
AND wo.wo_type = 'CORRECTIVE'
AND wo.opened_date >= DATEADD(year, -3, GETDATE())
ORDER BY wo.opened_date DESC;Reference architecture
The design is a reporting copy or API puller, a retrieval index over work orders and documents, a tool server of approved read actions, a local model served by vLLM or Ollama, identity mapping and an audit log. Use structured queries for counts and compliance numbers, and retrieval for questions like what fixed this symptom last time.
Quality of codes matters. If problem and cause codes are inconsistent, the model can still read the free text, but metrics such as MTBF, MTTR and PM compliance should be computed in SQL from clean fields. SMRP-style metric definitions and ISO 14224 style failure taxonomies help you decide what to compute.
In an air-gapped plant, the index, model and logs stay inside the enclave. Manuals are ingested from controlled media, and updates to the stack are brought in as reviewed packages.
Permissions, write-back and honest limits
Mirror CMMS security: site, department, craft and asset access. A technician should not retrieve cost data or work orders for a restricted area unless the CMMS already lets them. Carry these attributes into the index as metadata and enforce them in the tool server before any result reaches the model.
Write-back should use the supported API or the user interface, with a planner approving. Typical candidates are a draft corrective work order with description, asset, priority and suggested parts. Do not let the assistant close work orders, change PM intervals or adjust inventory without a person reviewing the change.
Limits: sparse or poorly written notes limit answers, and 'replaced part' as the only text teaches the model little. If your CMMS vendor ships its own assistant that works within your policy, use it. A private layer earns its place when you need to combine maintenance history with ERP, historian and document data on your own servers.
Sizing and phased rollout
Work order text for a mid-sized site is small by AI standards: hundreds of thousands of records fit in an index measured in gigabytes. Model serving is sized by users and questions, often one GPU for planners and technicians at one or two sites.
Phase 1: read-only history search with cited work orders. Phase 2: drafted work orders, failure summaries and PM review suggestions. Phase 3: approved creation of work orders through the supported interface. Measure with a set of past breakdowns where the planner knows the right answer.
Try it here, free
On-Prem CMMS Private AI Layer Blueprint Generator
Fill in your On-Prem CMMS environment and get a shareable architecture plan for a private, read-only AI layer: components, controls, rollout and open questions.
Common pitfalls
- !Reading production tables directly without vendor confirmation, then breaking on the next upgrade.
- !Ignoring site and craft security so technicians see cost or restricted areas.
- !Computing MTBF or PM compliance with the model instead of SQL on clean fields.
- !Indexing technician notes that contain names or personal details without redaction.
- !Letting the assistant close work orders or change PM intervals automatically.
- !Skipping manuals and OEM documents, which leaves answers without the procedure.
How an ERP-grounded AI assistant handles this
DataRay can query a CMMS reporting copy and file-share manuals on-prem and cite the work orders behind each answer. For maintenance-heavy sites, Netray can build the retrieval index, tool server and approval flow, and connect ERP and historian data so a planner sees cost, parts and condition in one answer.
Frequently asked questions
Does our CMMS need an API for this?
It helps. A supported API or reporting views are the safest routes. If neither exists, ask the vendor about supported export options. A scheduled export into a reporting store is better than querying production tables unsupported.
Can the assistant create work orders?
It can draft them. A planner reviews the description, asset, priority and parts, then creates the work order through the CMMS interface or supported API. Direct unreviewed creation is not recommended.
How does it find similar past failures?
Work order descriptions and close-out notes are embedded into a local index with asset class and failure codes as metadata. A question retrieves the closest records and the model summarises them, citing the work order numbers.
What about cloud-hosted CMMS tools?
The design still works if you pull data through the vendor API into your own store, but you then rely on the vendor for data access. Fiix, for instance, is SaaS with an API, not a database you host.
Where is the data stored?
In your reporting database and local index. The model runs on your GPUs, and logs stay on your servers. Nothing has to be sent to an outside AI service.
Related
How to Design Scrap and Downtime Reason Codes for MES and ERP
Keep one governed list of reason codes with a small hierarchy, owned in a single system and synchronised to the other. Separate scrap reasons from downtime reasons, mark each downtime code planned or unplanned and each scrap code as process, material, design or operator related, and map codes to ERP accounts or cost categories where accounting needs them. Fewer, clearer codes beat long lists nobody uses correctly.
How-toHow to Calculate OEE from MES Data and ERP Standards
OEE equals Availability times Performance times Quality. Availability comes from MES run and stop events against planned production time, Performance compares actual output to the ideal cycle time (usually the routing standard from the ERP), and Quality is good units divided by total units. The numbers are only trustworthy if planned time, ideal cycle time and the good/scrap split are defined once and used consistently.
AdvancedHow to add AI to GE Proficy (Historian and Plant Applications) on-prem
Keep Proficy where it is and put a read-only AI layer beside it. The assistant reaches Historian through its documented OLE DB provider or client SDK, reaches Plant Applications through its SQL Server database or a reporting copy, and runs an open-weight model on local GPUs. No tag data leaves the site, and every question and query is logged.
AdvancedHow to add a private AI layer to a custom MES on SQL Server
A custom MES has the advantage that you own the schema. Put a readable secondary or reporting copy beside it, expose approved views through a tool server, mirror your application roles with row-level security, and run an open-weight model on local GPUs. Write-back goes only through the MES application's own API with human approval.
AdvancedHow to add a private AI layer to Infor LX / BPCS on-prem
BPCS and Infor LX run on IBM i with their data in Db2 for i, in files that carry short coded names by module. A private AI layer reads those files through read-only SQL views over a replica, indexes documents and attachments separately, exposes approved queries through a tool server, and serves an open-weight model on local GPUs. Writes, if any, go through the supported transaction programs with human approval.
AdvancedHow to add a private AI layer to Infor System21 on-prem
System21 runs on IBM i with its business data in Db2 for i. The practical private AI pattern is a read-only replica or extract, SQL views that give files readable names, a retrieval index over documents, a tool server of approved queries, and a locally served open-weight model. User visibility is carried over from System21 security and company set-up, and writes only go through its supported entry points after approval.
AI for ERPAI 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.
AI for ERPAI for SAP Plant Maintenance: Notifications, Work Orders, and Predictive Signals
AI grounded on SAP PM data that triages notifications, drafts work orders, standardizes failure codes, and flags predictive maintenance signals, on-prem.
AI for ERPAI for IFS Maintenix in aviation MRO
AI over IFS Maintenix maintenance records, AD/SB compliance, and task cards for MRO, deployed on-prem so tech records never leave your network.
Stuck on On-prem CMMS (Maintenance Connection, MP2 and similar)?
Talk to engineers who work inside On-prem CMMS (Maintenance Connection, MP2 and similar) every week, and who build private AI that answers these questions from your own ERP data.
Keep exploring
Free tools
- On-Prem CMMS Private AI Layer Blueprint GeneratorFill in your On-Prem CMMS environment and get a shareable architecture plan for a private, read-only AI layer: components, controls, rollout and open questions.
- CMMS Overdue PM Work Orders SQL GeneratorFill in your table and column names and get a ready SQL query for overdue PM work orders and a PM compliance summary, in SQL Server and PostgreSQL syntax.
- On-Prem CMMS AI Assistant Sizing and Value CalculatorEstimate GPUs, index size, hours saved and payback for a private AI assistant over On-Prem CMMS.
ERP fixes and how-tos
- How to add a private AI layer to Infor EAM on-premReference architecture to add AI to Infor EAM on-prem: read-only data path, retrieval, permissions, local GPUs, write-back control and rollout.
- How to Add AI to IBM Maximo On-Prem Without CloudPrivate AI assistant architecture for Maximo Manage: REST object structures, security groups, DOCLINKS retrieval, local model, audit and approved write-back.
- How to add AI to Teamcenter on-prem without moving your PLM dataReference architecture for AI on Teamcenter on-prem: SOA and Active Workspace data paths, Access Manager permissions, local vLLM, audit and phased rollout.
Runs with your ERP
- Prometheus Planning & Scheduling with SAP Plant MaintenanceHow Prometheus Planning and Scheduling works with SAP Plant Maintenance work orders on S/4HANA and ECC: data flows, setup, pitfalls and a private AI layer.
- Oracle Fusion Cloud Maintenance with Fusion ERP and SCMHow Oracle Fusion Cloud Maintenance handles work orders, preventive maintenance and warranty on Fusion SCM: data flows, setup, pitfalls and private AI ideas.
AI on your ERP
- AI for UpKeep CMMS: Turn Work Order History Into Answers Without Handing It to a Public ModelUse AI with UpKeep CMMS through its REST API: question work orders, assets, parts and PM schedules, and tie maintenance data to your ERP and manuals.
- AI for Fiix CMMS: Ask Questions Across Work Orders, Assets and Parts Using Your Own Data BoundaryConnect a private AI layer to Fiix CMMS through its API: ask questions over work orders, assets, parts and meters, and link maintenance to your ERP.
Integrations
- How to add an AI chatbot to ProShop ERPHow to add an AI chatbot to ProShop ERP: a shop-floor and office assistant grounded in work orders, parts and quality records, with private model options.
- How to add AI agents to IFS CloudHow to add AI agents to IFS Cloud: a tool layer over OData projections, event-driven triggers, scoped service accounts and human approval before any write-back.
Software
- Accruent Maintenance ConnectionAccruent Maintenance Connection explained: CMMS work orders, preventive maintenance, assets, inventory, integrations, fit, limits and alternatives.
- Limble CMMSLimble CMMS explained: work orders, preventive maintenance, asset and parts tracking, API and integrations, fit for plants, limits and alternatives.