Source evidence
Sources and suggestions remain available so reviewers can see where values came from and spot ambiguities.
OptyWeb · Assisted automation
From email to a glass order, without losing control
AI Orders interprets incoming email requests, reads documents, identifies glass and processing, checks the ERP catalogue and applies reviewed customer rules. Your team validates exceptions before creating a Portal Order.
- Emails, PDFs and images
- Excel and CSV
- Customer rules
- Human review
How it works
Six steps from a message to an order ready for review
This is not a black box that invents part numbers: each stage handles a defined job and makes uncertainty visible.
- 1
Receive
The inbox collects requests and attachments for classification.
Organised intake
- 2
Classify
Separates orders from other messages before preparation.
Relevant requests
- 3
Extract
Reads text and document structure, including complex tables.
Source data
- 4
Interpret
Identifies lines, quantities, sizes, glass layers, spacers and processing.
Structured order
- 5
Cross-check
Looks up ERP items and applies approved customer rules when applicable.
Suggested references
- 6
Review and export
Staff resolve issues, confirm the Portal Order and can download its XML.
Controlled hand-off
Worked example
What happens when a customer writes 'as usual'?
The model recognises what is actually written. Customer conventions are handled by approved rules, while missing data stays open for review.
Subject: glass order
“Hello, we need 10 units of 4/16/4 as usual. Thanks.”
Quantity
10
Glass make-up
4/16/4
- Customer context
- Use a valid active rule only when it applies
- Needs human review
- Piece dimensions are missing: the system does not make them up.
What if a customer usually requests a special edge finish?
An administrator can configure a customer-specific rule. For instance, certain non-float glass types may require edge processing, with exceptions for small sizes. Without a valid rule, the system should not assume that work.
Illustrative, fictional example. It is not a measured result or a real interpretation screenshot.
What you can realistically expect
AI to interpret. Rules for context. People to decide.
Automation should not guess how each customer works. AI Orders combines initial document reading, company-specific rules and human oversight into one workflow.
Hypothetical example
- 01 · AI60–70 %Initial interpretation
Turning documents into data
Extracts information and suggests lines, sizes, glass make-ups, spacers and processing from the message and its files.
- 02 · RULES80–90 %Business knowledge
Customer-specific context
Cross-checks ERP catalogue items and applies approved rules. Corrections can generate suggestions for review.
- 03 · TEAMHuman oversight
Exceptions stay under control
A person confirms, corrects or holds uncertain details before an order is created.
Quality depends on source documents, ERP catalogue configuration and each company's rules. Accuracy must be measured with real examples.
Sources the module can read
Beyond PDFs: photos, emails and spreadsheets
The engine uses direct extraction for structured documents and visual AI for scanned or photographed content.
- TEXT
Emails
Plain text and HTML
Reads message bodies, subjects and relevant information to prepare an order request.
- PDF
PDF documents
Digital and scanned PDFs
Extracts embedded text directly, or uses visual analysis for image-based pages.
- IMAGE
Images and photographs
JPG · JPEG · PNG · WEBP
Can analyse order photos, tables, sketches and handwritten annotations. Uncertain readings need review.
- DATA
Spreadsheets and tables
XLSX · CSV
Reads rows and columns from modern Excel workbooks or CSV files, retaining the source structure.
These formats refer to the AI Orders extraction engine, not every file type OptyWeb can attach. Handwriting, sketches and scans depend on legibility and the configured AI model; uncertain results require human review.
Our method
Specialist AI plus verifiable business knowledge
The goal is not to claim 100% automation, but to reduce repetitive work while preserving technical judgement and review.
Customer-specific rules
Configuration can apply agreed conventions, items and exceptions without overriding what the customer explicitly states.
Supervised improvement
Human corrections can generate rule suggestions; activation is controlled, rather than blind automatic learning.
Product screenshots
How AI Orders fits into day-to-day operations
Real OptyWeb screenshots using commercial demonstration data. The interpretation example above is explanatory, not a real product screenshot.
01 · AI Orders inbox
Incoming requests, statuses and items requiring review before proceeding.
02 · Portal Orders
Once confirmed, the prepared order enters the Portal Orders flow for management and export.
03 · ERP orders
The ERP maintains its own orders and document tracking; XML export alone does not prove ERP acceptance.
Benefits
Less manual entry, more business judgement
Less retyping
Reduces copying order lines from emails and PDFs into forms.
Built for glass
Works with glass make-ups, spacers, thicknesses and processing, not just generic text.
Customer consistency
Applies reviewed conventions repeatedly without relying on staff memory.
Visible uncertainty
Missing or ambiguous data remains available for review before creating an order.
Connected to the ERP
Links review with Portal Orders, ERP catalogue items and XML export.
Company controls
Company-specific permissions and rules, with auditable changes.
Frequently asked questions
What to know before getting started
Optional module. Availability depends on each company's configuration and permissions.
Does it interpret 100% of orders correctly?
It would be misleading to promise that. Results depend on document quality and configured rules. Human review remains, and accuracy should be measured with representative real orders.
Does it learn customer habits by itself?
Corrections can lead to suggestions, but any rules affecting interpretation are reviewed and explicitly activated. It does not invent undefined customer conventions.
Does it create the order directly in the ERP?
After confirmation it creates a Portal Order and can export XML for the existing integration process. Import and actual acceptance in the ERP are separate steps.
What about complex files or incomplete details?
The module attempts to extract and interpret the content, but uncertain materials, unresolved items and missing fields require human review.
OptyWeb · AI Orders
Try it with documents representative of your business
A demonstration using your typical orders is the best way to evaluate rules, exceptions and potential time savings.