Improve AI draft order results
Give DMNexa clearer source text, maintain useful AI Brain rules, and review the details that language models most often misunderstand.
Use recent, explicit source text; separate customer facts from seller assumptions; maintain accurate AI Brain rules; and correct important fields during review instead of expecting perfect extraction.
AI works best when the input makes relationships explicit and your business rules stay current.
For typed or voice orders
State one fact at a time and label ambiguous details:
Customer: Hala. Item: one XL vegetable pizza. Add extra cheese, mushroom, and green pepper. Remove onion and olives. Pickup. Due tonight at 7 PM. Price not set.
Negative instructions deserve clear wording. Names, dates, numbers, addresses, and uncommon product terms deserve extra review after voice transcription.
For Instagram conversations
The assistant uses recent available context. If the final customer decision happened outside that context or in an unsupported attachment, add it during review. Do not rely on repeated re-analysis to invent missing information.
Keep AI Brain accurate
Open Settings → AI Brain and check:
- The business context summary.
- Answers imported from onboarding.
- Rules checked on every order.
- Custom instructions you added.
- Review strictness.
Remove contradictory or outdated rules. Write instructions as specific review needs, not broad claims such as “always make the order perfect.”
Correct instead of coaching the draft indirectly
When you know the right date, price, name, or item, edit the field. Re-analysis is useful when the source itself changed, but a direct correction is clearer and faster.
No configuration makes language analysis infallible. Keep seller review as the final quality check.
Frequently asked questions
Should I rewrite the whole Instagram conversation?
No. For a connected conversation, preserve the customer context and correct the resulting draft. For typed AI entry, provide a concise factual summary.
Will re-analyzing always fix a mistake?
No. Re-analysis can help when the source or context improves, but direct editing is more reliable for a known correction.
This page is maintained from the current DMNexa product behavior. If your screen looks different, check for an app update or contact support.