If you run a hotel F&B operation, you already know the invoice rhythm. Suppliers email PDFs (or print a docket on the delivery), someone types them into the stock system, someone else types them into Xero (or Sage / QuickBooks), and the numbers don’t quite match by the time the accountant looks at them six weeks later.
The drudgery of it is bad enough. The lag is worse. By the time you spot that beef is up 8%, you’ve already sold a quarter’s worth of margin-thin burgers.
What we changed
Inntally’s IntelliFlow module is built around three layers:
- Capture — suppliers send PDFs to
invoices@yourvenue.inntally.com, you snap on phone, or we auto-fetch from supplier portals. - Extract — AWS Textract reads the layout (OCR + table detection); GPT-4o-mini extracts SKU, quantity, unit, price, VAT, and matches to your standing catalogue.
- Validate + post — confidence-scored lines surface for human approval; once approved, the data hits your stock and your accounting in one click.
What our pilots are seeing (illustrative)
Across pilot hotels running IntelliFlow on their actual supplier mix, the headline numbers we see are:
- ~22 hours/month back to the F&B controller / GM (depending on invoice volume).
- 98% line-capture accuracy on clean PDFs; 90–94% on photographed dockets; lower on faded thermal print — flagged for review, not silently wrong.
- Cash position visible within hours of receipt rather than weeks.
- Supplier price drift (the silent margin killer) caught the day it lands.
These are illustrative ranges based on pilot data and industry benchmarks for Irish hotels — not promises. Your number depends on your supplier mix, your volume, and how dirty your invoice formats are.
The failure modes we’ve seen (be honest with yourself)
AI isn’t magic. The places IntelliFlow needs the most help are predictable:
- Hand-written delivery dockets in dim cold-storage lighting. OCR struggles; we flag low-confidence rather than guess.
- Supplier SKUs that don’t match anything in your catalogue. First couple of invoices need human mapping; thereafter automatic.
- Multi-rate VAT on one document. Common in mixed grocery + alcohol orders. Textract + our rules handle it, but we audit it for the first 2–3 invoices per supplier.
The pattern: AI does the boring 95% of the work; humans review the edge cases, which become rarer as the system learns your suppliers + catalogue.
What this changes about your week
Monday-morning bookkeeping doesn’t exist anymore. Your GM gets a phone notification that the day’s invoices are ready for approval — usually a 90-second screen. Your accountant sees the invoices in Xero within hours, not at the end of the month. Your supplier-cost drift is a daily nudge, not a quarterly horror.
Setting up
For new Inntally pilots, IntelliFlow is live in week one. We map your top suppliers to the catalogue, point invoices@ at the IntelliFlow inbox, and you start reviewing within days. Most of the gain is in the first 30 days; the rest comes from the cumulative effect of accurate live food cost feeding the recipe engine, the POS, and the analytics.
Read next
- IntelliFlow product page — the technical detail.
- Recipe costing that actually works — what to do once your costs are live.
- Security page — how we handle AI sub-processors under GDPR.
“Illustrative scenarios based on industry benchmarks and our pilot rollouts. Named case studies available under NDA on request.”