From copy-paste to connected in four weeks
A century-old Australian brand moved its customer service from spreadsheets to an AI-assisted service desk.
- Services
- Advisory · AI rollout · Integration build
- Period
- Jun – Jul 2026
- Team
- Eight staff across sales, marketing, finance and operations
Customer service ran on one inbox and one spreadsheet
Aurora Lites owns Redheads, the Australian matches and firelighters brand. The company has eight staff. Every customer contact, whether web form, phone, email or social media, landed in a shared inbox and was handled by one person.
Each case was manual. Copy the message into a chatbot with the customer experience guide, paste the reply back, ask for photos and batch codes, save the images, send voucher requests to finance, and log the case in a Google Sheet. Voicemails were typed out by hand. Five different AI tools were in use across the team, with no shared instructions.
The spreadsheet was the only record, so there was no reporting: no themes, no sentiment, no response times. The person doing the work was going on leave in August and someone else had to be able to take over. The brief asked for practical, working outcomes rather than simply an assessment
.
One AI tool, one record, four weeks
Discovery was one ninety-minute session plus access to the inbox, the tracker, the experience guide and the image archive. The rest of the time went into building.
The recommendation at the end of week one had five parts. Claude as the single AI tool for the business. Ticketing switched on in the helpdesk they already paid for. n8n as the integration layer. Airtable as a system of record the company owns. Training, with support through the first month. Owning the record was the priority, because it is the part that survives a change of vendor.
Every message becomes a case
Each message to the shared inbox runs through one pipeline. Rules handle the obvious cases. Claude classifies the rest. A case is created with the full conversation attached and stays current as the conversation continues.
- Intake
- Rules first, then classification by case type, sentiment and product. Marketing and supplier mail is filtered out before classification.
- Record
- Airtable replaced the spreadsheet. Replies, comments, status changes and closures sync back in real time.
- Images
- Complaint photos are read automatically. The batch code is extracted and posted back into the conversation.
- Voicemail
- Recordings are transcribed. Caller name, number and email are extracted and attached to the conversation.
- Reporting
- A live dashboard over every case: volume by month, channel mix, response and resolution times, sentiment at close, and goodwill spend.
The team runs it themselves
Everyone moved to one AI platform, with two training sessions and a customer service skill built from the company's own experience guide.
Five AI tools became one, with shared instructions.
Twelve months of history migrated into the new record.
Live on Claude in week three. Record and sync live in week four, before the August leave.
Estimated running cost of the new stack, against a one-off implementation fee.
Response time, resolution time and sentiment are now measured. The migrated history gives a baseline for the next twelve months.
Nicholas PsaltiSales & Marketing, Aurora LitesIt was an absolute pleasure working with Steve. We've built things above and beyond what we had hoped for, and we've really transformed the way we do business.
Have a problem shaped like this one?
Fixed-scope discovery to start.
Strata: Gerson Repreza — Unsplash