Published at: 2026-09-17
Smart order agent
The smart order agent turns unstructured procurement materials into structured transaction records.
It fits the execution stage after transaction confirmation and helps your team convert contracts, screenshots, Excel orders, and Chat Histories into CRM transaction objects faster.
Overview
Product positioning
The smart order agent is an AI assistant for sales, sales operations, and order handling scenarios.
It focuses on one core problem: customer materials are already available, so how do you turn them into orders, contracts, or other transaction records quickly.
It is not a simple OCR tool.
Its core value is to convert procurement details from contracts, screenshots, Excel files, and chat records into structured transaction results and then generate follow-up business records through your object model and field rules.
Core concepts
Material parsing
Material parsing means the system identifies procurement intent and business fields from text, Word files, Excel files, PDFs, screenshots, and images.Field mapping
Field mapping means the system maps the identified result to order, contract, and detail object fields based on your company object model.Transaction record generation
Transaction record generation means the system converts the confirmed result into transaction object records such as orders, contracts, and quotes.Industry know-how
Know-How is the set of company rules that covers field retention, material interpretation, special customer logic, and exception extraction standards.
It determines what the system should focus on, how it maps fields, and when it asks for human confirmation.Typical business scenarios
| Scenario | Typical question | Best-fit task |
|---|---|---|
| Attachment-based order entry | A customer sends a contract or PO. How do you enter it fast? | Identify the content and generate an order or contract. |
| Multilingual material handling | Overseas order materials are hard to read or slow to process. | Recognize and assist with multilingual order content. |
| Chat screenshot to transaction record | Customer requirements are scattered across chat screenshots or emails. | Consolidate the key details and convert them into a record. |
| Inconsistent templates | Customer materials arrive in many different formats. | Adapt to company fields and transaction structures dynamically. |
| Downstream transaction workflow | The record still needs approval, reminders, or conversion. | Continue into the downstream workflow after generation. |
Scenarios that do not fit
Do not start with the smart order agent in these scenarios:
- The deal is still in the inquiry stage and you only need pricing judgment.
- You only want simple reporting and do not need to generate a transaction record.
- The issue is account management rather than transaction record handling.
When the main question is how the price should be quoted, use the smart quote agent first.
Core value
The execution stage often includes these problems:
- Customer materials are scattered, so manual extraction is expensive.
- Orders and contracts contain many fields, so manual entry often misses details.
- Different customers use different material formats, which makes standardization hard.
- After order entry, the team still needs to handle approval and downstream workflow.
The smart order agent creates value in four ways:
- Lower order entry cost: It reduces manual input and material organization.
- Faster processing: It turns unstructured materials into system records faster.
- Fewer errors: It reduces missing fields and incorrect entry through identification and validation.
- Smoother downstream workflow: It lets orders, contracts, approvals, and conversions continue naturally.
Why industry know-how matters
The recognition quality of the smart order agent depends on your company-specific industry and business rules.
Companies differ in these areas:
- Which fields must be retained
- Which special product or customer rules apply
- How order objects and order line item objects are designed
- Which business signals are critical in contracts, POs, or screenshots
Because of that, the agent does not process every material set with one fixed template.
The system combines your object model, field design, and know-how to decide the extraction logic and output format dynamically.
Quick start
Learning goals
- Upload one unstructured procurement material set
- Identify and confirm the key business fields
- Generate an order, contract, or related transaction record
Prerequisites
- You have access to the smart order agent.
- You prepared a contract, screenshot, Excel order, PO, or another procurement material set.
- An admin already confirmed the key fields and required business extension rules.
Shortest path
- Open the smart order agent.
- Upload the procurement materials.
- Review the system result.
- Choose whether to generate an order, contract, or another target record.
Completion signs
- The system identifies the main fields.
- You can move into the record confirmation and submission page.
User guide
How to start order generation
You can create a new conversation in
ShareAgent and choose the smart order agent from the agent entry point.
You can also upload the materials directly and enter a command for transaction record generation so that the system routes the request to the right agent.Common prompts include:
- Generate a sales order from this contract.
- Identify this procurement attachment first.
- Generate a sales contract based on the attachment.
- Can we create an order directly from the requirement in this screenshot?
Typical workflow
Step 1: Provide the source materials
Upload the contract, screenshot, Excel file, PDF, or other procurement materials.
If the customer information is split across multiple attachments, upload them together.
The source manual highlights that the system supports online preview of the attachment content and page count after upload. This helps you confirm that the material set is complete.
Step 2: Review the recognition result
The system tries to identify products, quantities, customer information, and delivery requirements.
If similar or uncertain details exist, the system asks for confirmation through multiple turns.
Before you generate the formal record, you can also ask the system to View Attachment details first and confirm that the recognition result is correct.
This step is especially important when the source materials come from chat screenshots, scanned contracts, or multiple attachments.
Step 3: Confirm and generate the transaction record
After the recognition result is confirmed, choose whether to generate an order, sales contract, or another transaction object.
The system carries the result into the target page so that you can make a final review or edit.
If your company already customized the page or extended fields, that confirmation page still inherits the existing logic.
After submission, the system returns a Created Successfully result and supports copy, reminders, approvals, and downstream conversion.
The source manual also lists common conversion paths:
- Opportunity → Quote
- Quote → Sales contract
- Quote → Sales order
- Sales contract → Sales order
How to read the output
A useful recognition result should answer these questions:
- Which key fields the system identified
- Which details still require human confirmation
- Which transaction record type fits the current material set
- Whether you can move directly into edit and submission
How the result is reused
The smart order agent mainly turns recognition results into formal business records and Product Details on the target record.
Common reuse patterns include:
- Preview summary details in the recognition card.
- Confirm detailed fields in the new or edit page.
- Reuse the structured result in approval or downstream workflow.
Admin guide
Configuration prerequisites
Admins need permissions for object management, field maintenance, and business rule extension.
If your company uses standard fields and standard objects, you can start quickly.
If your company has special retention requirements for orders, contracts, or line items, complete the field design first.
What to prepare
Field design
If key details from the uploaded materials must be stored in order, contract, or Order Product objects, confirm first that the required fields already exist.
If not, add them to the target objects.
The source manual uses invoice title as an example and recommends adding the required field to the order or Order Product object first.
Industry know-how or special extraction rules
If your company uses special product extraction rules, special customer rules, or special downstream actions, extend the logic through the Skill Center.
Typical extensions include:
- Special field extraction rules
- Special customer business rules
- Post-generation workflow logic
The source manual uses this entry path: Agents > Skill Center > Create Skill.
Admins can import a skill or let AI create one to add special extraction logic and downstream actions.
Validation after configuration
Use a few typical but low-risk material sets for testing.
Check these points:
- Whether the key fields are identified correctly
- Whether the order or contract objects receive the result correctly
- Whether special rules are triggered as expected
- Whether downstream workflow works after the transaction record is created
FAQ
Which material formats are supported
The smart order agent supports text, Word files, Excel files, PDFs, images, screenshots, and voice materials.
How does it work with existing transaction record customization
The system returns the recognition result first and then opens the new or edit page for the target record.
Because of that, your existing page customization and field logic can continue to handle the final confirmation step.
Why does some information still need human confirmation
When the material set contains similar content, ambiguous wording, or missing information, the system asks for clarification.
This is a necessary control to reduce order entry risk.
When should you add company-specific know-how
Add company-specific know-how when you have special fields, special customer rules, special extraction logic, or special downstream workflow requirements.