Made2Manage + private AI
AI for Aptean Made2Manage: A Private Assistant on Top of Your M2M Data
Short answer
Aptean Made2Manage runs on SQL Server with a fairly open schema, which makes it a good candidate for a grounded AI layer: a private model reads order, job, inventory, and MRP tables through read-only views and answers the questions your Crystal Reports and ad hoc SQL currently exist to answer, without sending shop data to a public AI service. This suits the small and mid-size make-to-order and mixed-mode manufacturers who run M2M and cannot justify a large BI or AI vendor engagement.
- ERP
- Aptean Made2Manage
- Industries
- Discrete Manufacturing, Job Shops, Mixed-Mode Manufacturing
- Written for
- Operations Manager
Aptean Made2Manage (M2M) has been the backbone system for thousands of small and mid-size make-to-order and mixed-mode manufacturers for decades: order processing, engineering (BOM and routing), shop floor control, inventory, purchasing, and MRP all sitting on a SQL Server database with a schema that most long-tenured M2M shops know reasonably well, because they have had to. That familiarity is also the problem. The people who can answer 'what is the status of job 48213' or 'which open orders are short a component' quickly are usually one or two power users who know the tables and can write a query or a Crystal Report, and everyone else waits.
M2M shops tend to run lean on IT. There is rarely a dedicated BI team, and Aptean's own AI roadmap is oriented toward its larger cloud products, not M2M specifically. That leaves a real gap between what an M2M user needs, a fast, plain-language answer grounded in the actual job, order, and inventory data, and what is currently available, which is a saved report, an export to Excel, or a phone call to the person who knows the query.
The addressable pain is concrete: order status and shortage questions that come up dozens of times a day on the shop floor and in customer service, MRP exception review that a planner works through manually every morning, and job costing questions that come up at month end when someone has to explain a variance. None of that requires replacing M2M or moving to a different ERP. It requires a way to ask the SQL Server database a question in plain language and get an answer that is actually correct, with the source data shown alongside it.
Because M2M's database structure is well documented among its own implementation partners and not deeply obfuscated, standing up a grounded AI layer on top of it is a tractable project rather than a multi-year platform migration. The model runs on infrastructure the company controls, reads the SQL Server tables through read-only views, and never needs the underlying data to leave the building.
What usually gets in the way
The problems we hear most from operations manager teams running Aptean Made2Manage.
Order and job status knowledge concentrated in a few people
Answering a customer's 'where is my order' question quickly depends on someone who knows the M2M order, job, and shop floor tables well enough to query them directly; when that person is out, the answer takes longer or requires guessing from the screen.
Crystal Reports and SSRS reports lag what people actually need
Standard M2M reports cover the common cases, but a one-off question, such as which customers have orders affected by a specific late component, usually means a new report request that sits in an IT queue for days or weeks.
MRP exception review is a manual daily grind
Planners typically work through MRP action messages and exception reports line by line each morning, deciding which ones matter today, a task that is repetitive enough to automate the triage step without automating the planning decision itself.
Thin IT bench for a modern AI project
Most M2M shops do not have in-house data engineers or AI staff, so any AI initiative has to be scoped and delivered by people who already understand the M2M schema, not treated as a greenfield platform project.
No path to AI inside Aptean's own M2M roadmap
Aptean's newer AI features are concentrated in its larger cloud-based ERPs, leaving M2M customers without a vendor-native AI option unless they add a layer on top of the system they already run.
Where AI earns its place in Aptean Made2Manage
Each use case names the ERP objects it reads or writes, so your ERP team can judge the integration effort before anyone commits budget.
Order and job status lookup
Answers a plain-language question about a specific sales order or job's current status, referencing the order line, job routing operations completed, and open shortages, instead of a user navigating several M2M screens.
Touches: Sales order and order line tables, job/work order and routing operation records, inventory shortage data
Outcome: cuts the time to answer a routine order status question from several minutes of screen navigation to a direct answer
Shortage and component availability triage
Surfaces which open jobs are blocked on a missing or short component today, pulling from inventory and purchasing data, so a buyer or planner does not have to cross-check availability job by job.
Touches: Inventory on-hand and allocation records, open purchase orders, job material requirements
Outcome: planners resolve the daily shortage list in a fraction of the time spent checking availability manually
MRP exception summarization
Groups and prioritizes the MRP action messages that actually need a planner's attention today, filtering out the noise of routine reschedule suggestions from the ones tied to a real customer commitment.
Touches: MRP action messages, planned order records, sales order due dates
Outcome: planners work through the real exception list faster and catch fewer late orders slipping through routine noise
Purchase order and supplier follow-up drafting
Drafts routine follow-up messages to suppliers on open purchase orders approaching or past their promise date, for a buyer to review and send, rather than writing each one from scratch.
Touches: Open purchase order records, vendor master, promise dates
Outcome: buyers spend follow-up time on the suppliers that need real escalation, not routine status chasing
Job cost variance explanation
At month end, helps a controller understand why a specific job's actual cost diverged from standard by pulling the labor, material, and overhead postings behind the variance into a readable summary.
Touches: Job cost transactions, labor and material postings, standard cost records
Outcome: cuts the time to prepare a variance explanation for a specific job from an afternoon of digging to a focused review
Natural-language questions across order, inventory, and job data
Lets a customer service rep or manager ask a question once, such as which customers have orders at risk this week, and get an answer grounded in the actual M2M tables, without knowing the schema or writing SQL.
Touches: Order, job, inventory, and shipment tables accessed via read-only SQL Server views
Outcome: fewer ad hoc report requests land on the one or two people who know the M2M database well
New-hire and cross-training support
Gives a newer customer service or planning employee a way to ask how a process works or where a specific piece of information lives in M2M, reducing dependence on the one or two tenured staff who currently answer those questions.
Touches: Indexed internal procedure documents alongside live M2M query results
Outcome: new staff reach useful independence faster, and tenured staff spend less time on repeat basic questions
Reference architecture
The architecture treats the M2M SQL Server database as the system of record, reads it through dedicated read-only views rather than direct table access, and keeps the model and its logs on infrastructure the company controls given how small most M2M shops' IT teams are.
- 1
M2M database connector
Read-only SQL Server views over the order, job, inventory, purchasing, and MRP tables, built to isolate the AI layer from any direct write access to the production schema.
- 2
Data and semantic layer
A plain-language mapping of M2M's table and field names to the terms a customer service rep or planner actually uses, plus an index over existing Crystal Reports and any internal procedure documents.
- 3
Model serving
An open-weight model (Llama, Qwen, Mistral, or Gemma class) served with vLLM or Ollama on a modest GPU footprint the company owns or a small private cloud instance, sized to the query volume of a small or mid-size shop.
- 4
Retrieval and agents
Retrieval-augmented generation grounds every answer in current M2M data; any agent proposing a write, such as a draft supplier follow-up email, stops for a human to review and send it.
- 5
Governance and audit
Query and response logging tied to the user's M2M login, giving a simple, reviewable record even without a dedicated compliance function.
Integration notes for your ERP team
- Connects to the M2M SQL Server database through dedicated read-only views rather than direct production table access, isolating the AI layer from the live transaction path.
- Existing Crystal Reports and SSRS report definitions can be indexed as reference material so the assistant understands what a given report already covers before answering a similar ad hoc question.
- Where M2M's own API or COM-based integration points exist, they are used for any write-back action rather than direct database writes, preserving M2M's built-in business logic and validation.
- A lightweight identity mapping ties each user's M2M login to what the assistant can see, avoiding a separate access model to maintain.
- Works alongside existing ODBC-based reporting tools rather than replacing them; the assistant handles the ad hoc, plain-language questions those tools were never meant to cover.
- Sized for M2M's typical deployment scale: a single GPU server or a small private cloud instance is usually enough, not an enterprise cluster.
Deployment options
Small on-prem footprint
Shops that already run their M2M SQL Server on-site and want to keep the AI layer on the same local network
A single modest GPU server alongside the existing SQL Server instance, sized for the query volume of a shop this size rather than an enterprise-scale cluster.
Private cloud instance
Shops without server room space or in-house hardware management capacity
A dedicated, non-shared private instance that avoids the capital cost of GPU hardware while keeping the same read-only, human-approval architecture.
Hybrid with a hosted M2M instance
Shops that already have M2M hosted by a partner or on a private cloud
The AI layer sits alongside the hosted M2M database, connecting through the same secured network path the hosting partner already provides.
Compliance and data control
How the architecture supports your obligations. Certification and accountability stay with your organisation; the design keeps the evidence straightforward.
Customer and supplier data confidentiality
Read-only access and infrastructure the company controls keep order, pricing, and customer data inside the same boundary M2M already protects, rather than adding a new SaaS vendor with access to it.
ITAR / export-controlled data (where applicable)
For M2M shops that supply defense or aerospace customers, the same on-prem or private-instance architecture avoids sending technical data or order details to a public AI API, supporting the shop's own export control obligations.
Segregation of duties
The assistant's access mirrors each user's existing M2M security group, so a customer service rep cannot see cost data they could not already see in M2M, and any write suggestion still requires a human with the right role to execute it.
Audit trail for management review
Logged queries and responses give a manager a simple way to see what questions are actually being asked and answered, useful for spotting recurring reporting gaps without a formal audit program.
Where Netray fits
Custom build
M2M is not one of ERPray's current out-of-box connectors, so a purpose-built connector over the M2M SQL Server schema, using the same read-only, retrieval-grounded architecture, is the realistic starting point.
DataRay
Where a shop's procedure documents, quality records, or supplier files sit outside M2M as file shares, DataRay extends the same on-prem approach to that unstructured content alongside the M2M data.
How an engagement runs
Phase 1 . 1-2 weeks
Discovery
- -Review of the M2M database version, customizations, and existing Crystal Reports/SSRS reports
- -Use case shortlist ranked by how often the question comes up and how much manual effort it currently takes
- -Read-only view design covering the order, job, inventory, and MRP tables needed for the shortlist
- -GPU or private-instance sizing estimate scaled to shop size
Phase 2 . 4-6 weeks
Pilot
- -One or two use cases live in read-only mode for a defined group of users
- -Model evaluation against real M2M order, job, and inventory data
- -Feedback loop with the customer service, planning, or shop floor staff using it
- -Draft data flow documentation for management review
Phase 3 . 4-6 weeks
Production
- -Hardened deployment with access mapped to existing M2M security groups
- -Any approved write-back actions routed through M2M's own API or COM interfaces
- -Runbook covering model updates and basic monitoring for a small IT team
- -Short training session for the initial user group
Phase 4 . Ongoing
Scale
- -Additional use cases added from the original shortlist as they prove out
- -Rollout to additional departments (purchasing, customer service, planning) as adoption grows
- -Periodic review of which questions are being asked to spot new automation opportunities
- -Evaluation of newer open-weight models as they become available
Questions to ask any vendor, including us
A short list that separates real Aptean Made2Manage AI work from a chatbot demo.
- Does the vendor have direct experience with the M2M SQL Server schema, or will schema mapping start from zero?
- Does any part of the pipeline send order, customer, or pricing data to a public AI API by default?
- Can the assistant's access be scoped to match our existing M2M security groups?
- How does a draft supplier email or other write suggestion reach a human for approval before anything is sent or posted?
- What GPU or infrastructure footprint does this actually require for a shop our size?
- Can we see the underlying M2M data behind any answer the assistant gives?
- What happens to our data and model configuration if we end the engagement?
Frequently asked questions
Does Aptean offer AI directly for Made2Manage?
Aptean's publicly visible AI investment is concentrated in its larger, newer cloud ERP products, not specifically in M2M. M2M customers who want AI today generally add a layer on top of their existing SQL Server database rather than waiting on a vendor-native feature for this specific product.
Do we need to move to a newer Aptean product to get AI value?
No. Because M2M's schema is reasonably well understood and documented among its implementation community, a grounded AI layer can be built directly on top of the existing SQL Server database. This does not require migrating to a different Aptean product or ERP platform.
Is our M2M data safe if we add an AI layer?
It can be, if the model and the data it reads stay on infrastructure your company controls or a dedicated private instance, with read-only access by default and no default call to a public AI API. The architecture should be built so order, customer, and pricing data never leaves your boundary as part of normal operation.
How big a project is this for a small M2M shop?
Smaller than most people expect. A pilot covering one or two high-value use cases, such as order status lookup and MRP exception triage, is realistic in a matter of weeks, on a GPU footprint scaled to the shop's actual query volume rather than an enterprise deployment.
Can this replace our Crystal Reports and SSRS reports?
It is not meant to. Standard, recurring reports still belong in Crystal Reports or SSRS. The AI layer is built for the ad hoc, plain-language questions that do not justify a new formal report, and it can reference existing report definitions to stay consistent with how the business already measures things.
What is the realistic first use case for an M2M shop?
Order and job status lookup, and MRP exception triage, tend to show value fastest, since both are high-frequency questions with a clear existing process the assistant speeds up rather than replaces. Shortage triage is a close second once inventory and purchasing data is mapped.
Does this work if we have heavily customized our M2M installation?
Yes, though customizations need to be accounted for during discovery. Custom fields, added tables, or modified workflows just need to be mapped into the read-only views and the plain-language layer the same way the standard schema is, which is part of why a proper discovery phase matters before a pilot starts.
Related guides
AI for Aptean Intuitive ERP: A Private Assistant Without a Platform Change
On-prem AI for Aptean Intuitive ERP: a private assistant grounded in your Intuitive database for order, engineering, and shop floor questions.
Ask your ERP anythingNatural Language Query for ERP Data: Ask SAP, Infor, or Oracle a Question in Plain English
See how natural language query over SAP, Infor, Oracle, and NetSuite data works: grounded text-to-SQL, role-based permissions, and a visible audit trail.
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.
ERP AI Buyer GuideHow to Choose an ERP AI Implementation Partner
A CIO checklist for picking an ERP AI implementation partner: the architecture questions to ask, red flags, pricing models, and what to demand in the SOW.
ERP AI Cost GuideWhat ERP AI Actually Costs: A CFO's Guide
A CFO's guide to what ERP AI actually costs: GPU hardware, model licensing, integration and connector work, and realistic ongoing run-rate ranges.
QAD + on-prem AIAI for QAD Adaptive ERP in automotive and industrial manufacturing
Add AI to QAD Adaptive ERP or Enterprise Edition for automotive and industrial manufacturing, grounded on QXtend and QAD's API layer, on-prem or private cloud.
Plan it with numbers
Job Shop ERP Fit Assessment
Score how well your ERP handles high-mix, low-volume and make-to-order manufacturing across quoting, one-off routings, WIP visibility, subcontract tracking, and job-level cost variance.
Free ToolOn-Prem AI ROI Calculator
Turn hours saved per employee into annual net benefit, payback months, and 3-year ROI for an on-prem AI investment.
Free ToolERP Data Quality for AI Assessment
Score item master, customer, vendor, and transaction data quality to find out whether your ERP is ready to ground an AI copilot, forecast, or chatbot.
GuideNatural Language ERP Query Interface
Query your ERP using natural language. Transform plain English questions into SQL/API calls with LLM-powered interfaces that democratize ERP data access.
GuideSelecting an AI Implementation Partner: Evaluation Criteria
Selecting an AI implementation partner: a weighted evaluation scorecard, reference-check questions, and why a pilot-first contract beats a big-bang one.
GuideMaster Data Management (MDM) Strategy for ERP Systems
Implement a master data management strategy for your ERP. Establish golden records, governance policies, and data stewardship for lasting data quality.
Talk it through with an engineer who knows Aptean Made2Manage
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.