Three people, 700 storefronts, and a human approving every reply
A Sydney founder, a Hamburg CEO, a UK founder and a US AI consultant on who should answer the customer when AI technically can.
Three people run customer support across roughly 700 storefronts.
In the UK, one founder is the help desk.
Opposite setups, one shared worry. As a Sydney founder puts it, "someone has to own it if it's wrong."
We asked which parts of a job AI has actually taken and which still need a person. Here the job is customer support, and the answers change with the country, the product and the cost of one bad reply.
The AI reads and drafts. A person signs off.
David Caruso, Founder & CEO of BuyFactory.direct (Sydney, Australia)
“Our customer support AI reads every incoming message and writes the first reply against the real order. That's why three people can run support across roughly 700 storefronts.”
“A person still reviews and approves every reply before it goes out, not because the AI can't technically send it, but because someone has to own it if it's wrong.”
The machine does the reading and the first draft. The human does the owning.
Triage went to the bots. Trust took months to rebuild.
Matthias Woggon, CEO & Co-Founder of Eyefactive GmbH (Hamburg, Germany)
“First-pass support triage moved to AI almost completely at Eyefactive. Tickets about our AppSuite Touchscreen CMS are now sorted, tagged, and matched to known fixes before anyone on the team reads them.”
“An expired price ran in dozens of stores for a whole weekend. The data took an hour to fix. Rebuilding trust with store managers took months.”
Speeding up the sort is easy. The expensive failures are the ones that reach a store.
The help desk has a founder on the other end.
Jake Wardle, Founder of EV Cable Hub (Wetheral, United Kingdom)
“The help desk runs Monday to Friday, 8am to 6pm UK time, and the founder is the person on the other end.”
No debate about automation here. Support is a promise, and he keeps it himself.
Give AI 97%. Customer service is where she holds the line.
Goddess Wealthy, (GoddessWealthy.com) Certified AI Consultant (United States)
“I would give AI 97% of the work and have humans handle the 3%, the stuff that only we can do as humans.”
“Customer service is where I absolutely want to keep humans. I myself have felt very frustrated when I call my bank and it's an AI program that does not understand 90% of what I'm saying, and it refuses to connect me with a human.”
The most AI-forward number in this piece still stops at the support line.
Where they split
Sydney: AI reads and drafts, and a person approves every reply before it goes out.
Hamburg: AI sorts, tags and matches tickets, and people own the moments where a wrong answer costs money.
The UK: the founder answers the help desk himself, Monday to Friday.
The US: give AI 97% of the work, but keep humans on customer service.
The overlap: for all four, the customer-facing moment is where a person belongs.
My take
Nobody here is anti-AI. Reading, sorting and drafting have moved to machines, and good riddance. The real line sits where a wrong answer reaches a customer. Draw that line first, then pick the tool.
Your turn
Where in your support flow does a human have to say yes before a customer sees the result? Readers and featured experts, tell us where your line sits.
Next: in US and Australian real estate and UK aerospace, AI fully takes over data entry and search, but a wrong closing date or a safety sign-off shows why the final call stays human.
