AI Assistant

Ask anything about your plant, and stay in control of every change

Several AI surfaces connected to different parts of VECHR MES, rather than one chatbot for everything. All of them follow the same principle: the AI proposes, a person confirms, and nothing in production data changes without approval.

In short

Ask questions about your plant in plain language. The AI drafts routings, flows and master data, and a person confirms every change.

Today Answers wait for someone to build a report.

With VECHR MES Ask in plain language. Get the answer from live data.

Today Configuration means navigating many forms.

With VECHR MES The AI drafts it. You review and confirm.

Today AI cost is a black box.

With VECHR MES Token use is tracked per user, per model, per turn.

Five surfaces, one principle

Each is designed for different work, but all of them hold the same line: AI helps you make the decision, it does not take the decision over.

Surface	What it does
AI Insight	Answers everyday-language questions about KPIs, OEE (Overall Equipment Effectiveness), downtime, and the status of a particular batch or lot.
MES Advisor	Helps answer configuration and usage questions from the product documentation, rather than generating an answer from general knowledge.
MES Expert Engineer	Drafts master data changes for a person to review. It never executes those changes itself.
Orchestrator Assistant	Turns a plain-language description of an automation into a draft flow that can be reviewed before deployment.
Routing Designer Assistant	Turns a plain-language description of a production process into a draft routing to review before use.

This boundary is part of the system design: a proposed change has to pass human confirmation before it is applied or saved as a change.

Ask directly, without opening a report

Ask about a batch in plain language Find status, quantity, current resource, and execution information without writing SQL or building a custom report first.
Answers come from production data Questions about the operation are answered from the MES records available, so the context follows the production conditions the system is actually recording.
It stays advisory AI can spot anomalies, explain conditions, and make recommendations. The decision and the action stay with the user.

Predictions are computed, not guessed

AI Insight opens from the top bar (⌘⇧A or Ctrl+Shift+A). Behind it are three specialised read-only agents for performance, production and quality, so each conversation loads only the tools it needs.

Six predictive and prescriptive tools Machine health prediction, maintenance priority, production completion prediction, material shortage prediction, quality trend analysis and schedule optimisation suggestions.
Calculated from MES data These tools compute their results directly from the records in the MES. The figures are not estimated by the language model.

AI costs you can see

Measured per user, model, and turn Input and output token usage is recorded for every interaction, so AI consumption can be measured clearly.
Filterable and reviewable Usage can be viewed by user, model, and time range, so AI cost has clear context and a clear owner.
Can be enabled separately AI Insight and AI Agent run as their own microservices, so their use can be separated from the core MES deployment.
Choose your provider Run on Anthropic Claude or OpenAI GPT and pick the model. The API key stays on the server, a provider can be tested before switching, and a mock provider is available for testing.

Frequently Asked Questions

What executives and IT usually ask before enabling AI.

Can the AI change our production data on its own?

Not directly. The surfaces that produce changes only create drafts, and a user has to review and confirm before anything is applied. That boundary is part of the system design, not a permission that can be loosened.

Where do the answers come from?

Questions about the operation are answered from the MES data and records available. MES Advisor uses the product documentation as its context for configuration and system usage questions.

How is AI cost controlled?

Token usage is tracked per user, model, and turn, and can be filtered by time range. Administrators can see usage patterns and AI consumption clearly.