Automation That Pays For Itself

⚠️The Problem

Our client — a 12-person building materials wholesaler based in Sydney — was processing 200–300 supplier invoices per month entirely by hand. Their accounts manager spent two full days every week opening PDF invoices, manually typing figures into Xero, matching them against purchase orders in a spreadsheet, and chasing approvals by email. The process was slow, error-prone, and completely dependent on one person.

Three invoices had been paid twice in 12 months due to data entry errors — costing the business over $4,100 in recoverable losses and hours in reconciliation.

⚙️What We Built

• 1Built an OCR pipeline that reads incoming PDF invoices from their shared inbox, extracts supplier name, invoice number, line items, totals, and GST automatically
• 2Connected the pipeline to Xero via API — invoices are created in draft automatically, matched against existing purchase orders
• 3Approval workflow built in Slack: Sarah receives a Slack message with the invoice details and clicks Approve or Query — no more email chains
• 4Exception handling for invoices that don't match a PO — flagged for human review only
• 5Dashboard showing invoice processing status, approval queue, and monthly volume

Tech Stack

📈The Results

⚠️The Problem

Our client — a dental group operating three locations in the UK — had reception desks that were overwhelmed. Each location was fielding 60–80 calls per day — the majority being appointment booking requests, "what are your opening hours?", and "can I reschedule?" questions. Receptionists had no time to focus on patients in front of them.

On top of that, no-shows were running at 18% — each missed appointment cost the practice £80–£150 in lost revenue and an unfilled chair. Reminder calls were supposed to happen but were frequently skipped due to workload.

⚙️What We Built

• 1Deployed an AI chatbot on all three clinic websites handling booking, rescheduling, FAQs, and new patient intake
• 2Integrated directly with their practice management system (Dentally) via API — appointments booked by the bot appear in the calendar in real time
• 3Automated reminder sequence: SMS 48 hours before → Email 24 hours before → SMS 2 hours before, with a one-tap confirm/reschedule link
• 4New patient intake form automated: chatbot collects medical history, insurance info, and preferences before the first visit — pre-populated into Dentally
• 5After-hours coverage: chatbot handles all enquiries 24/7, escalating urgent dental emergencies to an on-call number

Tech Stack

📈The Results