AI document & data processing

From scattered documents
to data ready to process.

AI reads documents, extracts the information you need, categorizes and checks it. Staff do less data entry, know exactly what needs verifying, and data goes straight into the process you already use.

Already in the SAL software Discuss your needs Xem demo
Illustrative demo · Simulated data
Invoice (sample)
No.: HD-0001231
Date: 15/09/20262
Seller: Sample Co., Ltd.3
Total: 12,540,000 VND4
1Invoice numberHD-000123
2Date15/09/2026
3SellerSample Co., Ltd.
4Total12,540,000 VNDNeeds checking
Already in the SAL software

The SAL software already covers data

  • Import from Excel / CSV into any section, read right in the browser; export any list to Excel.
  • E-invoices from MISA, Bkav and SePay connect directly; download all of your business data with one button.
  • The assistant adds fields, tables and workflows from a description, no programming needed.

The service below is for reading paper invoices, forms and specialised documents into the software; SAL - AI sets it up inside your own software.

Start with the Free plan See the software-building assistant
02 · The problem

Reduce time spent on manual reading, entry and checking.

Invoices, forms, emails and reports come from many sources in different layouts. Staff have to find information, re-enter it and check it before moving on. We help businesses organize these steps into a trackable process.

Re-entering information

Data is copied from documents into spreadsheets or software.

Inconsistent formats

Each source uses different naming, structure and layout.

Exceptions are hard to spot

Missing, conflicting or duplicate information slows down the next step.

03 · Capabilities

How does AI help process documents?

Intake and categorization

Receives documents from supported sources and groups them by record type or business rules.

Information extraction

Reads the required fields from PDFs, scans, forms and documents within the tested scope.

Data standardization

Converts dates, field names and values into a consistent format following your business rules.

Checking and flagging

Detects missing fields, invalid values or data that needs verifying before use.

Cross-checking information

Compares data in documents against your reference sources, using agreed criteria.

Review and data export

Sends unusual cases to the responsible person. Results are only exported or updated in the system after being checked.

Document types, data fields and integrations are confirmed through assessment and testing on real samples.

04 · Demo

From one input document to one verifiable record.

Illustrative scenario: processing incoming invoices.

Illustrative demo · Simulated data

Original document

Invoice (sample)
No.: HD-0001231
Date: 15/09/20262
Seller: Sample Co., Ltd.3
Total: 12,540,000 VND4

Extraction result

1Invoice numberHD-000123
2Date15/09/2026
3SellerSample Co., Ltd.
4Total12,540,000 VNDNeeds checking
View details of the field to check

Field 4 (Total): 12,540,000 VND differs from the sum of line items, 12,450,000 VND. The system flags it and points to the location in the original document so staff can confirm or correct it.

  1. 1

    Read and extract

    The system captures the invoice number, date, seller details and amounts according to the agreed fields.

  2. 2

    Check against rules

    Flags missing fields or discrepancies that need verifying, with their location in the original document.

  3. 3

    Review and transfer data

    Staff check the flagged fields before exporting or updating the software.

Book a process demo
05 · Use cases we can deploy

Which of your tasks are a good fit?

Invoices and vouchers

Extract data, check required fields and prepare for system entry.

Purchase orders and delivery notes

Capture item codes, quantities and dates and cross-check them with reference sources.

Forms and records

Categorize, check for missing information and forward to the responsible person.

Reports and data tables

Compile information into an output template, noting sources and items to check.

Each project starts with a specific group of documents. For contracts or specialist documents, AI assists with extraction; the responsible person still confirms the conclusions.

06 · Deployment process

Three stages, from data to operation.

01

Define the data to process

Review sample documents, fields to capture, validation rules and the system receiving the data.

Output: Scope, evaluation sample set and acceptance criteria.

02

Build and test

Set up extraction, standardization, checking and review routing; evaluate on representative samples.

Output: Pilot process and results for each document type.

03

Operate and improve

Controlled rollout, error monitoring and adjustment when formats or business rules change.

Output: An operating process with the agreed support scope.

07 · Handover and measuring results

What do you get?

  • List of supported document types and fields.
  • Rules for standardization, checking and review routing.
  • An interface or method for checking results.
  • Confirmed integrations and export formats.
  • Test sample set with acceptance results.
  • User guide and maintenance scope.

How are results measured?

  • Accuracy per data field.
  • Processing time from input to approved result.
  • Share of documents requiring staff intervention.
  • Effort spent checking and correcting errors.
  • Cost per completed document.

We measure the time spent correcting errors too, not just how fast the AI reads documents.

08 · Fit and data control

This solution fits when…

The business processes repetitive documents, knows exactly which fields to capture and has someone to confirm results. The value of deployment depends on volume, complexity and current handling effort. Blurry, handwritten or highly variable documents need testing before the scope is confirmed.

Every result must be traceable to its source.

Every extracted value links back to its location in the original document so staff can check it. Who has access, where data is stored, for how long and which provider's technology is used are clarified before deployment.

09 · FAQ

Frequently asked questions

Can it handle PDFs and scanned images?

It can be deployed with supported formats. Image quality, language and layout affect results; the scope is confirmed through testing.

Is every document read accurately?

No. Accuracy varies by document type and by field, so the system is evaluated separately for each group and important information always goes through a checking step.

Can it connect to accounting software or internal systems?

It depends on whether the software supports integration (API) and the access you can grant. If a direct connection isn't possible, data can be exported to a file for import into the software.

What happens when document templates change?

New formats need to be evaluated and the process may need adjusting. The update scope is agreed in the support package.

What does the business need to prepare?

Representative sample documents, a list of fields to capture, validation rules, an output template and a person responsible for business requirements.

How is the cost calculated?

The quote separates deployment, integration, usage and maintenance. Document types, page count and processing volume are the factors to assess.

10 · Contact

Which documents take your team the most time?

Tell us the document types, processing volume and the data you need to receive. We'll work with you to define a suitable pilot scope.

You don't need to upload documents at this step. Sample documents will be exchanged later via an agreed method.

The information you provide is handled under our Privacy policy.

Your request has been received.

Thank you. We will get back to you to discuss a suitable scope.