It is a Tuesday. Marco runs an online eyewear shop based in Italy. Orders come in 24 hours a day. Customers send messages constantly: "When will my glasses arrive?" "Does this frame fit my face?" "Can I return these?" "Is the prescription right?"
Before they brought in AI, the average customer waited 5 minutes for a reply. After hours, they waited until the next morning. Cart abandonment was high. Pre-sale questions died on the vine.
In 2023 they brought in two boring tools: a chat widget on their website, and an AI assistant trained on their own product catalogue. Nothing custom. Nothing built in 2 weeks for $1,500. Just two off-the-shelf tools, configured carefully, and reviewed before launch.
By the end of the year, the company — Eye-oo, an Italian multi-brand eyewear platform — had added €177,000 in extra revenue and reduced average customer wait time from 5 minutes to 30 seconds. A 25% sales lift.
A note on these numbers: they are reported by Tidio, the vendor of the chat tool Eye-oo used. Vendor-reported numbers should always be read with caution — the vendor has every reason to make their tool look good. But Eye-oo is a real business with a real catalogue you can browse today, and the pattern they describe is consistent with what other small businesses report when they use the same approach.
This post is about that approach.
The number that should make every small business owner pay attention
98%.
That is the average open rate of a message sent on WhatsApp, according to Meta's own WhatsApp Business data. For comparison: a marketing email opens at around 20%. SMS opens around 90%. The 98% figure was confirmed again in 2026 by Twilio's Messaging Engagement Benchmark Report, which analysed 4.8 billion WhatsApp messages across 62,000 business accounts and found a 98.2% open rate.
WhatsApp dominates because users check it dozens of times a day and treat it like a personal channel, not a marketing channel.
In Australia, WhatsApp is now the preferred channel for many customer-supplier conversations: tradies and clients, salons and clients, accountants and clients. Australian small businesses increasingly run their entire customer comms on WhatsApp Business — the free app from Meta — and never look at email again.
The problem: as soon as you have more than 20-30 active customer conversations, the messages stop being a strength and start being a liability. You miss replies. You forget to follow up. You answer "what's your address?" 50 times a week. You feel guilty. The customer feels ignored.
This is exactly the gap where boring AI fits.
What Eye-oo actually did
Eye-oo runs a website with a chat widget. When a customer arrives or starts asking questions, the chat opens. Behind the chat, an AI assistant — called Lyro, made by Tidio — answers based on Eye-oo's own product catalogue and FAQs.
The AI does not invent answers. It is trained on Eye-oo's data — return policy, shipping times, prescription rules, frame sizes, colour options. When a customer asks "can I return these glasses?", the AI gives the actual return policy. Not a guess. Not a generic answer.
When the AI is not sure, it hands off to a human. When a customer wants to talk to a real person, the AI asks them to wait and a human takes over.
This is not "AI replacing customer service." It is "AI handling the boring 80%, humans handling the complicated 20%."
Out of 2,233 customer conversations in their reported period, the AI fully handled 1,825 — about 82%. Humans took the remaining 408. The AI also captured 1,305 leads — people who started a conversation, gave their email, and could be followed up later.
The cart-abandonment piece is the part that matters most for any e-commerce business. When a customer added items to their cart and then left without buying, the chat sent a friendly follow-up. From 537 of those follow-up conversations, the system contributed an estimated €1,600 in sales in 2023.
The full €177,000 number combines all of these effects: faster sales conversations, better cart recovery, more after-hours support, fewer lost leads.
Why this works for an Australian salon, café, dentist, or trade business
Eye-oo is e-commerce. But the recipe is not e-commerce-specific. Substitute the artefacts:
- A salon's "prescription rules" become "treatment options and pricing"
- A café's "frame sizes" become "menu items and dietary information"
- A dentist's "shipping times" become "appointment availability and prep instructions"
- A tradie's "return policy" becomes "service area and call-out fees"
The pattern is the same: a chat widget on the website, or a WhatsApp Business number, or both. An AI trained on the business's own information. Boring questions answered automatically. Complicated questions handed to the owner. Quote follow-ups, appointment reminders, after-hours acknowledgements all sent automatically.
The recipe in 5 steps
What Eye-oo did right — and what any Australian small business can copy:
1. Started with one real problem: waiting times.
Not "let us transform our business with AI." Just: customers wait too long. AI fixes that one thing.
2. Used an AI tool trained on their own data, not generic AI.
Lyro was given Eye-oo's product catalogue, FAQs, and return policy. It answers from those, not from random ChatGPT-style output that might invent prices or rules.
3. Reviewed the AI answers before going live.
Eye-oo did not just turn the AI on and hope. They tested the responses, refined the training data, and watched the early conversations.
4. Kept humans in the loop.
The AI handles 80%. Humans handle the 20% the AI is not confident about. Customers are never trapped in a robot loop with no escape.
5. Measured what changed.
Conversations handled. Average response time. Leads captured. Cart-recovery sales. Without measurement, "AI is helping" becomes guesswork. With measurement, you know whether it is worth keeping.
This is the opposite of vibe coding. This is boring AI that works.
What you can do this week
You do not need a developer. You do not need a $1,500 agency build. You do not need 2 weeks of agency time.
You need:
- A WhatsApp Business account (free, set up in 30 minutes)
- One AI chat tool (most have free tiers — Lyro, Intercom Fin, ManyChat, and others)
- One hour to load your FAQ, opening hours, services, and prices into the tool
- One week of close watching to make sure the answers are accurate
- The willingness to hand off to a human when the AI is not sure
Start with one channel. Start with one type of question — "what time do you open?", "how much for X?", "how do I book?" — and let the AI handle just that. Add more questions over time. Stop when something does not work and fix it before adding more.
This is what 40% of Australian small businesses are quietly doing right now. They are growing 2.8x faster than the rest. Not because they vibe-coded a custom app. Because they started boringly.