What DMNexa AI can—and cannot—do with an order conversation
Understand how DMNexa analyzes messages, why details can be missing or wrong, and what sellers should verify before approving an AI-created draft.
DMNexa AI can extract and organize details that appear in a conversation, identify gaps, and apply seller-defined review rules. It cannot guarantee intent, invent reliable missing facts, or replace the seller’s final review.
AI is good at turning language into a possible structure. It is not a recording of objective truth. The most reliable way to use it is to understand both sides of that boundary.
What DMNexa AI can do
Depending on the conversation, DMNexa can:
- Recognize items, quantities, sizes, variants, and customizations.
- Preserve free-form instructions that do not fit a catalog option.
- Detect fulfillment language such as pickup or delivery.
- Extract dates, addresses, customer details, and stated prices.
- Compare the draft with the rules in your AI Brain.
- Identify missing information and potential conflicts.
- Suggest what you may need to ask next.
What it cannot guarantee
AI cannot guarantee that:
- A customer’s ambiguous phrase has only one meaning.
- The latest message did not reverse an earlier instruction.
- A name from the social profile is the correct order name.
- A product or price exists unless your connected data supports it.
- A default date is the date the customer wanted.
- An address, allergy note, or custom option is complete.
It also cannot decide whether you should accept the order. Capacity, safety, policy, and commercial judgment remain yours.
How to review efficiently
Do not reread every field with equal effort. Focus first on details that are expensive to get wrong:
- The requested product and quantity.
- “No,” “without,” “instead,” and changed instructions.
- Due date and time.
- Pickup versus delivery and the full address.
- Personalization, allergies, dimensions, or event requirements.
- Price and any money already received.
Use the warning panel and missing-field list as a checklist, then compare important claims with the source conversation.
A good AI result can still be incomplete
If the customer did not provide a name or delivery method, the correct AI behavior is often to leave it missing. A polished-looking guess would be worse. Fill the field after the customer answers, or edit it when you already know the information from a reliable source.
For product steps, see Review and edit an AI draft and Improve AI draft results.
Frequently asked questions
Why does an AI draft sometimes show missing fields?
The conversation may not contain those details, or the wording may not be clear enough to map confidently. Missing fields are a review aid, not necessarily an error.
Why can a date appear even when the customer did not provide one?
An editable order may use a visible default date. The review warning should distinguish that default from a date extracted from customer intent, and the seller must verify it.
Should I approve an order when the AI confidence looks high?
Review it anyway. Confidence is an aid, not a guarantee, and negative instructions, dates, addresses, and custom details deserve special attention.
This page is maintained from the current DMNexa product behavior. If your screen looks different, check for an app update or contact support.