Questions Plant Managers Actually Ask
The daily questions of a converting plant are specific. What is on the schedule for press 2 tomorrow and is material ready? Which jobs are past due and what are they waiting on?
How much of last week's film was wasted and on which jobs? Which customers have most reprints? What did we ship yesterday and what is still on the dock?
Each of these is answerable from the MIS or ERP, but only if someone builds the report, and the report often needs data from several tables or from both the MIS and the accounting system. Plant managers end up exporting to Excel or waiting for IT.
- Late jobs by cause: material, artwork, tooling, capacity
- Waste and makeready by press, substrate and shift
- Margin and reprints by customer and job type
- Inventory of rolls and sheets against upcoming demand
- On-time delivery by customer and product family
How a Question Layer Works
A question layer translates a plain question into a query against your database or API, runs it with the asking user's permissions, and returns the result with the query visible. That visibility matters: the plant manager can verify that the answer used the right date range and the right definition of late, and IT can review the logic. It should be read only.
Netray's ERPray is built along these lines for ERP data, and the same approach applies to an MIS database. It does not change the system, create jobs or move inventory.
Make Sure the Data Is Answerable
A question layer reveals data quality problems fast. If waste is not recorded per job, the layer cannot report it. If reason codes for late jobs are free text, answers will be messy.
If the same customer exists under three names, margin by customer will be split. Before starting, list ten questions you want answered and check whether the fields exist and are populated. Where they are not, fix capture first: required fields at job close, controlled reason code lists, and customer master cleanup.
These improvements help every report, not only AI.
Try it here, free
Natural Language ERP Reporting Savings Calculator
Turn report volume, build time, and automation rate into the monthly hours and dollars saved by letting users ask ERP questions in plain English instead of building reports.
Permissions and Trust
Packaging companies hold customer pricing and confidential specs. The question layer must respect existing ERP or MIS roles, so a press operator cannot ask for customer margins. Keep the model on your servers or in a private cloud if customer agreements restrict data sharing.
Log questions and queries. Also set expectations: the layer gives answers as accurate as the data and the definitions, and a wrong definition of on-time delivery will produce confident but misleading numbers. Agree on definitions with the team before rollout.
When a Standard Report Is Better
If a question is asked every day and the answer must be identical every time, build a standard report or dashboard in the MIS or a BI tool. Daily production summaries, shipping lists and the board of the day should be fixed, tested reports. A question layer is for the ad hoc question that nobody built a report for, and for people who would not write SQL.
If your team already has good dashboards that people use, adding a chat interface will not improve much and may add noise.
Frequently asked questions
Can I ask my print MIS questions in plain English?
Not natively in most products, but a separate layer can read the MIS database or API and translate plain questions into queries, with results and the query shown. It needs read access, a map of your data and agreement on definitions such as late and waste.
Will AI reporting work if my job data is incomplete?
It can only report what is recorded. Missing waste, time or reason codes will limit answers, and inconsistent customer names will split totals. Use the first questions to find gaps, then fix capture at the shop floor and in master data.
Is it safe to connect AI to a production MIS?
It can be if the connection is read only, uses a limited database account or API role, honors user permissions, and the model runs inside your own environment. Test on a copy of the database first and review the logs of generated queries.
Give Netray ten questions your plant manager wishes the MIS could answer and we will show which are answerable today from your data, with the query shown.