# How to add a private AI layer to Infor EAM on-prem > Source: https://netray.co/erp-help/private-ai-layer-infor-eam-on-prem-architecture > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: ERP Help ## Short answer On-prem Infor EAM stores its data in an Oracle or SQL Server schema whose tables use the R5 prefix, and it exposes web services for supported access. A private AI layer reads equipment, work order and parts data through read-only views on a reporting copy, indexes failure notes and manuals for retrieval, mirrors EAM organisation security, and serves an open-weight model locally. Work orders created or changed by the assistant go through EAM's web services after a person approves. ## Kind advanced ## System on-prem-apps ## Erp Infor EAM (on-prem) ## Module Private AI / Integration Architecture ## Query how to add AI to Infor EAM on-prem ## Alternate Queries - Infor EAM AI integration - Infor EAM work order chatbot - local LLM for maintenance history EAM ## Applies To Infor EAM on-prem, formerly Datastream 7i, on Oracle or SQL Server; table details should be verified against your version. ## Steps ### Reference build for Infor EAM **Items:** - List the entities behind your first use cases: equipment, work orders, parts, stores, PMs. - Build read-only views with business names over the R5 tables on a reporting copy. - Index manuals, procedures and work order notes with organisation and department metadata. - Serve an open-weight model on local GPUs and expose parameterised queries through a tool server. - Map each user to EAM organisations and departments and enforce them on queries and retrieval. - Add an audit log of question, query and documents retrieved. - After a read-only pilot, add drafted work orders submitted through EAM web services. ## Sections ### Data access paths in Infor EAM **Body:** Infor EAM tables carry the R5 prefix. In commonly documented schemas, R5OBJECTS holds equipment and assets, R5EVENTS holds work orders, R5PARTS holds the parts catalogue, and R5STORES and stock tables hold inventory. Work orders carry columns such as EVT_CODE, EVT_DESC, EVT_OBJECT, EVT_STATUS, EVT_ORG and EVT_MRC (the responsible department). Confirm column names and any user-defined fields against your version's data dictionary before building views. The supported route for transactions is EAM's web services, which provide SOAP and REST style access to entities such as work orders and equipment; your version's documentation defines the exact services. ION connectivity is available in some deployments for event-driven integration. For reading, a reporting copy of the database is simpler and does not load the application tier. **Code:** -- Open work orders with equipment and department (verify columns in your version) SELECT e.EVT_CODE, e.EVT_DESC, e.EVT_OBJECT, o.OBJ_DESC, e.EVT_STATUS, e.EVT_ORG, e.EVT_MRC FROM R5EVENTS e LEFT JOIN R5OBJECTS o ON o.OBJ_CODE = e.EVT_OBJECT WHERE e.EVT_STATUS NOT IN ('C','CAN') -- adjust to your status codes AND e.EVT_ORG = :org ORDER BY e.EVT_CODE DESC; ### Reference architecture, including the air-gapped option **Body:** A workable private AI layer for Infor EAM has six parts. First, a read-only data path: R5 table views and EAM web services over a replica or reporting copy so production is not queried. Second, a retrieval index over the documents and attachments people actually search, built with an open embedding model and stored in a vector database that runs on your servers. Third, a tool server, for example an MCP server, that exposes a short list of approved read actions with typed parameters. Fourth, an open-weight model served with vLLM or Ollama on local GPUs. Fifth, an identity layer that carries EAM user group, organisation and department into every request. Sixth, an audit log of who asked what, which queries ran and which documents were retrieved. The model never receives a database connection string. It asks the tool server to run a named action, and the tool server builds the parameterised query, applies the user's visibility filter and returns rows. This keeps free-form SQL generation out of production paths and makes every answer traceable to a query you can show the user. For an air-gapped site, all six parts run with no outbound route; model weights and updates arrive on approved media and are scanned and checked before use. Use a read-only reporting copy: an Oracle standby or read replica, or a SQL Server readable secondary or log-shipped copy. Keep the assistant off the primary so that maintenance planners and mobile technicians are never slowed down. Attachments and documents are usually stored outside the database tables, so index them from the document store or file share rather than from SQL. ### Permissions: the assistant must never show more than the user may see **Body:** EAM controls access by user group, organisation and department, and many screens filter by organisation. The tool server should identify the signed-in person, look up their organisations and departments from EAM, and add those predicates (for example on EVT_ORG and the equipment organisation column) to every query. Do not rely on the model to remember them. Documents attached to work orders and equipment may have narrower rights than the work order itself. Carry the access control list into the retrieval index as metadata and filter at query time. ### Write-back only through supported interfaces **Body:** If you want the assistant to draft work orders or add comments, make it prepare the payload, show it to the technician or planner, and submit it through EAM's web services under that user's credentials after approval. EAM then applies its own validation, workflow and audit trail. Never write to R5 tables directly. ### Sizing, phased rollout and honest limits **Body:** Sizing depends on concurrency, not on headcount. A pilot of 10 to 20 users with a 7B to 14B class model can run on one workstation-class or datacentre GPU with 24 to 48 GB of memory. A site with 100 to 300 named users, of whom a third use it daily, often needs one to four GPUs depending on model size and answer length. Retrieval indexes for tens of thousands of documents are typically in the single to low double digit gigabytes. Use the sizing calculator for your own numbers and treat results as ranges. Roll out in three phases. Phase one is read-only questions over a small set of views and documents, with every answer showing the query or sources. Phase two adds drafts, such as a summary, a message or a prepared transaction, that the user reviews but that nothing posts. Phase three adds approved actions through Infor EAM's supported interfaces, always under the approving user's identity, with an audit record. Free-text failure notes are inconsistent. Retrieval over them is useful but will surface noise, so test on real histories. Equipment hierarchies and custom fields differ widely between sites. If your main need is dashboards and standard maintenance KPIs, EAM's own reporting and analytics are enough. A model adds value for natural-language history search, cross-asset comparisons and document question answering. ## Pitfalls - Querying the production database during shift change when technicians are closing work orders. - Assuming status codes are the same in every organisation. - Indexing attachments without carrying their access rights. - Writing directly to R5 tables and bypassing workflow and validation. - Ignoring user-defined fields that hold the information users really search for. ## Ai Angle ERPray and Netray's custom build work follow this pattern: read-only by default, the query is shown to the user, Infor EAM roles and visibility are respected, and everything runs on your hardware with open-weight models. Where a vendor's own AI or standard reporting already answers the question, we say so; the private layer earns its place for cross-system questions, document search and sites that cannot send data off premises. ## Faq ### Is Infor EAM's own AI enough? **Answer:** If you are on a cloud-hosted EAM with vendor AI features and are comfortable with them, perhaps. For on-prem installs with data that must stay on site, a private layer lets you use open-weight models without moving data. ### Which database do you support? **Answer:** The pattern works with Oracle and SQL Server because it uses read-only views on a reporting copy. Syntax details differ and views are written per engine. ### Can technicians use it on mobile? **Answer:** A web front end served on your network can work on tablets. Offline use is a separate topic and not something we would promise by default. ### How is access limited by site? **Answer:** Organisation and department are resolved from EAM and applied as filters by the tool server on every query and retrieval request. ## Search terms Infor EAM AI, Infor EAM on-prem AI integration, Infor EAM work order chatbot, R5EVENTS R5OBJECTS query, Datastream 7i AI, maintenance history LLM, private AI CMMS on-prem, Infor EAM web services, local LLM maintenance data --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json