From patterns to understanding: a new way to work in eye-share Capture

With eye-share Capture’s new interpretation engine, you spend less time correcting documents and more time validating trusted results — without patterns or setup.

Let’s start with a simple challenge.

What does the handwritten text in red say?

Take a second look.

Is it A55? 455? Something else?

If you’re unsure, you’re not alone — and yet invoices like this arrive every day. That hesitation is exactly why document interpretation has always been harder than it looks — especially at scale.

Why real-world documents are hard

Invoices and business documents rarely arrive in neat, predictable formats. In reality, they come with:

• Different layouts and structures.

• Multiple languages and number systems.

• Local calendars and VAT rules.

• Handwriting, stamps, noise, and poor scan quality.

For years, automation has relied on OCR — technology that reads characters and turns them into text. That works well when documents are clean and consistent.

But finance doesn’t need text.

Finance needs meaning.

Where OCR reaches its limits

OCR can tell you what characters are on a page.

It can’t reliably tell you:

• Which number is the total.

• Which date actually matters.

• Whether a value is VAT, currency, or something else.

To compensate, systems have relied on patterns, templates, and rigid rules. You told the system where to look.

These work — until a document changes slightly. And real documents change all the time.

That’s where errors, exceptions, and manual corrections creep in.

The shift: from reading characters to understanding meaning

This is where the new interpretation engine changes the game. Instead of reading characters, it understands context, structure, and intent.

It interprets documents more like a human would — but consistently and at scale.

That means:

• Complex layouts are no longer a problem.

• Different languages and number formats are understood.

• Calendars and VAT structures are interpreted correctly.

• Even hard-to-read, unstructured documents are handled with confidence.

All without training, templates, or document-specific setup.

High accuracy from day one.

Not all numbers look the same

Here’s another example:

This number is perfectly valid — it’s just (hand)written in a different numeric system.

For international businesses, this is a real and recurring challenge. Invoices arrive from suppliers across regions, using different languages, number formats, calendars, and numeric systems. Validating whether a number is actually correct isn’t always straightforward — even for experienced finance teams.

This is where AI interpretation makes a difference.

Instead of relying on familiarity or predefined formats, the AI understands numbers in context — across languages and numeric systems — and interprets what the value represents within the document.

The result:

• Fewer validation errors.

• Less manual checking.

• Greater confidence when working across borders.

For global operations, this isn’t a nice-to-have. It’s essential.

And hey, the number is 2000 if you were wondering. We didn't get it right the first (second or third) time either, but the new eye-share Capture did. 

What this means in practice

For customers using Capture Services, the impact is already clear.

By combining intelligent interpretation with our specialist team, automation has increased dramatically — in optimised environments, automation levels went from 70% to approach 99%.

• Far fewer exceptions.

• Less manual correction.

• Faster processing.

• More trust in automation.

For customers running eye-share Capture on their own, the shift to AI still means significantly better handling of complex documents — with far less dependency on rigid setups.

Same documents. Better understanding.

From patterns to prompts

Another important change happens behind the scenes.

Earlier approaches depended heavily on predefined patterns and rules. These are fragile by nature — a small change in layout can break them.

With AI interpretation, instead of maintaining patterns, you guide the system through prompts.

Small layout changes no longer break interpretation.

This makes document handling:

• More resilient to variation.

• Easier to maintain.

• Better suited to real-world documents.

Less time spent maintaining patterns. More time spent actually processing invoices.

Same workflow. Better data.

One thing worth highlighting: The user interface stays the same. The difference is in what the system delivers.

To the left you see the "old" eye-share Capture, and to the right the new. 

The shift from OCR to AI happens behind the scenes. You still review, verify, and process documents in the same interface — just with better data and fewer corrections needed.

Same workflow. Smarter interpretation.

Understanding financial structure

Let’s look at one last example.

This is a structured invoice table — but structure alone doesn’t make interpretation easy.

Several numeric columns sit next to each other: unit price, amount, VAT, and total. The values are similar. The background is uneven. Grid lines are faint.

OCR can extract numbers. 

But finance doesn’t work with isolated numbers.

It works with relationships: which value is VAT, which is the subtotal, which includes tax, and how they connect. That requires context. 

You control the confidence

AI suggestions in eye-share Capture always come with a confidence score — and you decide what level is acceptable.

For example:

Set a higher threshold:

→ More automation.

→ Fewer documents stopped.

Set a lower threshold:

→ Stricter control.

→ More review.

This gives you full control over the balance between automation, accuracy, and risk.

The AI supports your process — it never removes control.

Back to the first question

Remember the red number at the start?

Correct answer: 455

The AI interprets it correctly. Not because it recognises a shape, but because it understands where the number appears, what it represents, and how it fits into the document as a whole.

That’s the real shift.

From reading characters - to understanding meaning.

A quiet upgrade — with a big impact

This isn’t a future promise or a beta feature. The AI interpreter is already in use, delivering results every day.

eye-share Capture still does what it’s always done:

• Captures documents.

• Interprets data.

• Validates accuracy.

• Sends clean data into your workflow.

But now, it does it with AI at the core — built on trusted, enterprise-grade technology from Microsoft.

From teaching the system where to look — to validating what it understands.

Same purpose. Much stronger foundation. A new way of working in eye-share Capture.

If you’re ready to reduce exceptions, eliminate pattern maintenance, and get more predictable automation in eye-share Capture, reach out to our team or book a demo.