The recommendation comes first
Every action starts with a material-specific instruction. Rows show the action needed, rather than asking the user to interpret a generic risk label.
PLATFORM / AI DECISION SUPPORT
An answer your team can act on.
Use generative AI to explain the evidence, compare practical alternatives and help your team decide. Keep the material’s calculated quantities and dates in view.
Decision supportOn demand
What should we do before this material runs short?
MUDAGEN
Recover the timing gap first.
- Check the existing receipt.
- Confirm the quantity needed early.
- Compare a feasible alternative.
EvidenceAlternativesDecision
Illustrative conversation · Reasoning stays connected to the supplied material evidence.
Specific actions. Visible reasoning.
A deeper review on request
Generate an AI recommendation when you need reasoning, alternatives or a second look. Opening the dashboard does not run a paid generation.
A conversation in context
Ask follow-up questions by typing or recording your voice. Review the transcription before sending, and choose spoken replies when useful.
Better context from your team’s experience.
When a buyer rejects an action, the reason matters: the supplier cannot split a delivery, a material is awaiting quality release, or a transfer would create another shortage. Relevant feedback can inform a later AI review.
This is controlled use of decision history, not automatic model retraining. The public demo keeps bounded feedback in the browser and lets the user inspect or disable it. Shared company memory, its access rules and retention are agreed in a customer implementation.
AI supports decisions within the supplied evidence. ERP execution follows the permissions and approval policy configured for your deployment.
Make AI usage a deliberate choice.
| Activity | How it works in the public experience |
|---|---|
| Browse materials and actions | Data-based guidance; no background generation |
| Generate an inventory recommendation | Explicit user action with limited reviewer access |
| Ask a text question | Explicit Send with bounded recent conversation context |
| Ask by voice | Record, stop, review transcript, then send |
What is calculated by the system and what does generative AI add?
The inventory logic produces data-based guidance such as coverage, timing gaps and proposed quantities. Generative AI explains the situation, compares eligible alternatives and helps a user explore the decision in conversation. It should preserve the evidence rather than invent a new stock balance or unsupported date.
Does AI learn when we reject a recommendation?
A recorded rejection reason can provide relevant context for a later review. In the public demo, that feedback is bounded, inspectable and stored in the browser; it is not autonomous model retraining. Shared company memory, permissions, expiry and evaluation are a separately scoped part of a production implementation.
Do visitors need ChatGPT accounts? Can they use voice?
Public website and demo chat do not require a ChatGPT account. Users can type or choose Voice, allow microphone access, stop recording and review the transcript before sending. Spoken replies use the browser’s speech service. Chat and voice use MudaGen’s API connection and are subject to request limits.
Will AI run for every material and create unexpected costs?
The public product keeps deeper AI recommendation generation manual. Dashboard navigation and spreadsheet upload do not trigger background recommendation batches. Text and voice chat are explicit actions with separate limits. For a company deployment, agree generation policies, volume assumptions and usage costs in the commercial scope.
YOUR NEXT MOVE
Let’s start with
your operation.
Bring a recurring inventory challenge. We’ll discuss the data, scope and a practical path forward.