Case study · Julz Shoes · Live since April 2026
The 24/7 agent.
Julz Shoes was receiving up to 150 WhatsApp queries a day from its customers. Today an always-on AI system answers nearly all of them in seconds, and seamlessly hands over to the human team when they're needed.
Where it started
150 customer queries a day
Julz Shoes is a South African leather footwear brand, which prides itself on a high level of customer service. In recent years its online side has grown fast, and with that growth came the increasing customer queries. Routine questions about sizing, stock availability, style advice, and way more.
On a normal day that came close to 150 messages. Two to three members of the Julz team carried most of it, and the line did not respect office hours, so the work often ran into the evenings.
The team did not want to lose the warmth their customers were used to. They wanted that warmth to stop costing them their day.
Watch JulzBot in action
A real shape of a conversation
See JulzBot live at julzshoes.com →
Most messages look like this. A question comes in, an answer comes back in seconds, and it ends before it ever becomes a task for anyone.
JulzBot answers from live store data and a knowledge base the Julz team keeps current, so a customer asking about sizing, stock, delivery or an order gets a real answer straight away — not a holding reply.
And it knows its limits. When something genuinely needs a person — like arranging an overnight delivery — it offers to hand over, then passes the whole thread across so the customer never has to start again. The Julz team picks up a conversation already in context.
A question answered in seconds, then a clean handover the moment a person is genuinely needed.
JulzBot is an automated assistant. AI can make mistakes.
What was built
JulzBot, in six parts
JulzBot is one always-on system made of six specialists that work in sequence. Each message is read, sent to whoever is best placed to handle it, answered from live store data and a knowledge base the Julz team controls, and escalated to a person only when it should be. Lighter steps like routing run on a faster, lower-cost AI model and the harder reasoning on a stronger one, which keeps the cost per conversation low.
The Gatekeeper
Reads every incoming message first and screens out anything harmful or off-topic before it goes any further.
The Router
Works out what the customer actually wants and hands the message to the right specialist.
The Stylist
Handles catalogue, sizing, fit and styling questions, and shares the right product links straight from the store.
The Policy specialist
Covers delivery, returns, exchanges and tracking, answering from the rules the Julz team sets and keeps current.
The Orders specialist
Pulls live order status so a customer asking "where is my order" gets a real answer, not a holding reply.
The Handover specialist
Recognises when a human is genuinely needed and brings one in, passing the full conversation across so the customer never repeats themselves.
It is connected to the Julz store for live product and order data, to their WhatsApp line for messaging and human escalation, and to a knowledge base the team can update themselves whenever a policy or product changes.
The first week measured
30 April – 7 May 2026
A single reporting week, taken straight from the system.
Conversations handled across the week
Customer messages answered
Resolved with no human involved (458 of 472)
Average turns to resolve — most done in under two exchanges
What changed for the team
From up to an hour to seconds
Before JulzBot, a customer might easily have waited up to an hour for a reply, and only during office hours. Now the answer arrives in seconds, at any time of day or night. The handful of cases that still need a person reach the Julz team already in context, so the human picks up a conversation rather than restarts one.
The Julz team didn't disappear from the line — they just got freed up for the parts of customer service that actually need a human.
That is the quiet handover. The everyday questions are caught and answered the moment they arrive, and the people who used to spend their day on them now spend it where their judgement actually counts.
What this means for your firm
The context was retail. The problem is not.
Julz Shoes is a fashion retailer, but the problem this system resolved is not unique to retail. Any business handling a consistent volume of inbound queries manually has the same opportunity — the specific questions look different in every firm, but the underlying situation is the same: a team answering the same things every day, when a system could be doing it instead.