Prepared for Esprigas · Feb 09, 2026

Your path to document automation starts here

Thank you for taking the time to explore Lido with us, Chris. Here's everything we discussed, personalized for your team's needs.

↓ Scroll to explore

01

What we heard

Outgrowing Nanonets after years of use
You've been with Nanonets a long time and they've done a great job, but you're outgrowing their capabilities, especially on nested tables, handwritten propane invoices, and complex document structures.
27,000+ documents per month and constant model retraining
20,000 invoices, 2,000 supplier statements, 5,000 customer POs. Even though 80% of your suppliers use the same two systems, you're still spending "a ton of time retraining models."
Manual approval exists only because you can't trust the extraction
You'd love to automate more, but you can't. Every document missing a required field goes to manual review — not because of business rules, but because of accuracy concerns.
02

How Lido solves this

Challenge
Nanonets can't handle nested tables, handwritten invoices, or combined PDFs
Our Solution
Lido's AI extraction with Vision OCR handles all three. We tested live on your worst documents
Process 100% of documents through one system instead of routing exceptions to manual handling
Challenge
Constant model retraining every time formats change or suppliers are added
Our Solution
Plain-language rules that generalize across all suppliers. Zero ML training
Eliminate retraining overhead; onboard new suppliers in minutes, not days
Challenge
Manual approval exists solely because extraction accuracy isn't trusted
Our Solution
Higher baseline accuracy + confidence scoring enables rules-based auto-approval
Shift from reviewing every document to exception-only review
03

Your projected ROI

Net annual savings
Here's how we got there
Current State
Invoices, customer POs, and supplier statements processed per month
Time per Invoices, customer POs, and supplier statement (manual verification)
Monthly hours on Invoices, customer POs, and supplier statement processing
Fully-loaded hourly cost
Monthly cost of manual processing
With Lido
Estimated time reduction
Hours saved per month
Monthly savings
Lido cost (estimated at your volume)
Net monthly savings
Hours returned to your team annually
Annual cost savings
04

Companies like yours

Disney Trucking

Freight & Logistics

Trucking company processing handwritten driver tickets across multiple locations — 6 full-time employees doing nothing but manual data entry.

Results:
  • 6 FTEs freed from manual data entry
  • 360,000 pages per year processed automatically
  • Handwritten driver tickets extracted with 100% accuracy
  • One template handles all document variations
"We have six full-time people processing these tickets. This is all they're doing."

Like you, they were processing difficult handwritten documents at scale. One template handles it all — no retraining required.

05

Questions you raised

?

How does API integration work with Business Central?

We can push directly to Business Central via API. You'd set up endpoints for each document type:

‍

  1. 810 for invoices
  2. 850 for POs
  3. 855 for order confirmations

‍

Once the first one is configured, additional endpoints take 15-30 minutes each. One customer set up 631 endpoints in a month — you won't need anywhere near that, but it shows the flexibility.

‍

This replaces your current workflow of:

  1. Exporting XML from Nanonets
  2. Routing through AS2 connection
  3. Processing through EDI software
  4. Manual approval gates
?

How do you handle nested tables with sub-segments, like cylinder groupings?

We showed this live across multiple calls. When a document has grouped items (like "RNT U510" containing nitrous oxide and CO2 cylinders underneath, or "CD" numbers for other equipment), we:

‍

  1. Extract each sub-item as its own line
  2. Use the beginning balance for split quantity
  3. Multiply unit price by quantity for each row
  4. Skip the parent row that's just a category header

‍

The rule is simple: "When you encounter RNT and a unit number, do not extract that row. Instead, split into the number of rows below it using their beginning balance."

‍

During your team demo, we extended this to CD numbers as well. Same logic, different prefix. One instruction handles both patterns.

‍

Your reaction: "That looks great."

?

Can we link back to original documents for verification?

Yes. Since you use OneDrive, we can store processed documents there and include the link in the extracted data. When a user sees something that looks wrong in Business Central, they click through to the original document instantly.

‍

This replaces your current AWS storage bucket workflow from Nanonets.

‍

06

Your investment

Recommended for your volume
$ /month

Ready to get started?

Most teams are live within a week.
Schedule a call