Studio Hugo

Case studies

Three problems I solve most often.

AI that reads documents, answers from your own systems, and works on numbers you can trust. Real projects, anonymised.

Wholesale

AI reads your documents. The order lands in the ERP.

Purchasing worked from paper lists and supplier PDFs. Now AI reads the condition sheets, the system suggests what to order and where it’s cheapest, and the order goes into the ERP in one click.

  • Purchasing orders most items through the cockpit and sees the saving per order
  • Order capture runs automatically
  • No direct write access to the ERP database

Challenge

  • Supplier condition sheets only as PDFs
  • Purchasing based on paper lists and experience
  • Online orders retyped by hand
  • ERP reachable only through a proprietary protocol

Approach

  • AI extracts conditions from supplier PDFs, a person checks them
  • Purchasing cockpit with order suggestions and the cheapest source
  • REST interface as the single write path into the ERP
  • Order importers with duplicate protection

Data protection: Writes only through one controlled interface. AI suggests, a person approves.

SystemsLegacy ERPSupplier documentsOnline partner channels
Leisure and tourism

Ask your systems. Get one answer.

Bookings, POS and invoicing lived in three systems, the P&L in Excel. Now management asks Claude about its numbers and gets one answer across all three, with sources.

  • Revenue reconciled to the cent with the tax advisor
  • EBITDA reconciled to the cent against the annual accounts
  • Reports in under 0.2 seconds

Challenge

  • Three systems without a shared data model
  • P&L assembled by hand from Excel exports
  • Nobody sure the numbers were right

Approach

  • Read-only inventory of all systems
  • Nightly reconciliation into a separate data layer
  • Management cockpit: P&L by business area, actual vs. plan
  • AI access through Claude

Data protection: Read-only access. Customers only as codes. Every access logged.

SystemsBooking systemPOSInvoicingERP
Wholesale and leisure

Numbers you can trust, before AI touches them.

AI is only as good as the data behind it. Before connecting anything, I get the numbers right and prove it.

  • Rebuilt rules match the existing classification for 98 to 99% of items
  • Every deviation explained
  • Revenue and EBITDA reconciled to the cent

Challenge

  • Item groups and prices had drifted from the business rules
  • Goods billed on use, with gaps nobody saw
  • Revenue in three systems that didn’t match

Approach

  • Business rules rebuilt from a legacy routine and checked against live data
  • Daily, complete consignment billing with traceable gaps
  • Reconciliation against the tax advisor’s figures and the annual accounts

Data protection: Every check runs read-only. Changes only after a dry run and review.

SystemsLegacy ERPBooking systemPOSAccounting

Selected work

More work, in production.

Client work is anonymised. The systems run every day.

Legacy ERP

One clean door into a closed ERP

I turned a proprietary ERP protocol into a REST API with roles, duplicate protection and staging. Every new connection goes through one controlled door, never straight into the database.

REST APILegacy ERP
Order intake

Orders that import themselves

Shop, marketplace and partner orders land in the ERP automatically and only once. Items that can’t ship get flagged instead of overlooked.

Automated importMonitoring
Purchasing

A purchasing cockpit with AI

It suggests what to order, how much and where it’s cheapest. AI reads the supplier condition sheets, and the order goes into the ERP in one click.

AI document readingERP
Master data

Business rules rebuilt from a legacy routine

Prices, flags and item groups had drifted because an old routine stopped running. I rebuilt the rules, checked them against live data and explained every deviation.

Data qualityReconciliation
Billing

Consignment billing without gaps

Goods sit with the customer and are billed when used. Now invoicing runs daily and in full, and every gap is traceable.

BillingReconciliation
Warehouse

Pick lists that only promise what’s on the shelf

Batch pick lists include only orders that can really be picked in full. A warehouse dashboard answers “busy or slow?” with numbers.

WarehouseDashboards
Reporting

Reports at the push of a button

Month-end, manufacturer reports and requests from authorities: one click instead of manual Excel work, delivered on schedule by email, to OneDrive or to Google Sheets.

Automated reports
Pricing

Prices that update themselves

Prices and margins that adjust on their own, based on your rules and live data from other systems.

PricingIntegration

Book a call

Your systems next?

15 minutes. We look at your systems and what you want from them. I’ll tell you straight whether and how I can help.

Book a 15-minute call →

or write to alan@hugo.studio