Before You Deploy AI Agents, Find Your Leaks
Every founder I speak to in Dubai this year has been asked the same question, by a supplier, a board member or their own team: what's your AI plan?
It's a fair question. In May 2026, Sheikh Hamdan bin Mohammed launched a two-year initiative to move Dubai's private sector towards agentic AI: systems that don't just answer questions but take actions, such as booking, following up, updating records and making decisions. Dubai Chambers is delivering it through training, incubators and dedicated funding.
The direction is clear, and the tools are getting better every month. But there's a step most businesses are about to skip.
Why SMEs feel the pressure
For a founder-led business, the pitch is attractive. An AI agent that answers every enquiry instantly, books appointments, chases quotes and never takes a day off.
Competitors are talking about it. Suppliers are selling it. It feels like falling behind if you don't.
So the temptation is to buy a tool, plug it into WhatsApp or the website, and switch it on.
That's where it goes wrong.
AI amplifies whatever it sits on
An AI agent doesn't invent a customer journey. It runs the one you already have, faster and at scale.
If your journey works, that's brilliant. If it has leaks, the agent will reproduce them: every time, for every customer, without anyone noticing.
If nobody follows up after an enquiry today, an agent that replies in two seconds still leads to nobody following up.
If your team doesn't know what's been promised, an agent that makes promises makes the problem bigger.
If there's no agreed next step after a first conversation, the agent won't know what to offer either.
I've spent more than 20 years in contact centres, and I've seen the same thing with every new technology: IVR menus, live chat, chatbots, CRMs. The businesses that got value from them had fixed the process first. The ones that didn't just made their problems faster and more expensive.
Four things to fix before you automate
1. Know what happens today
Map the journey as it actually runs, not as it's meant to. When someone gets in touch at 10pm on a Friday, what happens? Who sees it? When do they reply? If you can't answer that in a few sentences, you're not ready to hand it to an AI.
2. Give every stage an owner
Enquiries, follow-up, onboarding, renewals. Each one needs a named person. An AI agent can do the work, but someone has to be accountable for the result, and someone has to pick up when the agent hands over.
3. Decide what "good" looks like
How fast should an enquiry be answered? What should happen after a trial or a quote? What does a new customer hear in their first week? Write it down. An agent can only follow standards that exist.
4. Fix the handover to a person
Every AI agent will reach a point where a human needs to step in: a complaint, a complex question, a big decision. That handover is where most automated journeys break. Make sure the person picking it up has the context, and knows it's theirs.
Where AI genuinely helps once the basics work
None of this is an argument against AI. Once the journey works, it can make a real difference, especially for a small team:
Where it sits | Automate it once it works | Fix it first if… |
Enquiries | Instant replies to common questions, out-of-hours cover, booking links. | Nobody owns enquiries today, or replies take more than a day. |
Bookings and reminders | Confirmations, reminders, rescheduling. | Bookings regularly go missing or double up. |
Follow-up | Nudges after a trial, quote or visit, routed to the right person. | There's no agreed next step after first contact. |
Onboarding | Scheduled check-ins in the first seven days. | You haven't decided what a good first week looks like. |
Handover to a person | Passing the conversation, with context, to someone who can act. | Nobody is clearly responsible for picking it up. |
The pattern is simple. Automate what already works. Fix what doesn't before you automate it.
Don't buy the tool first
The most expensive mistake I see is a business that signs up for an AI platform and then works out what to use it for. You end up shaping your customer journey around the tool's features instead of around your customers.
Start with the journey. Find the leaks. Fix the ones that matter. Then choose the tool that fits.
The Leak Score as your pre-automation check
The YOURCXC Leak Score™ looks at the five points where revenue leaks: enquiry, first contact, onboarding, retention and referral. They're the same stages most AI agents are being sold to automate.
You get an instant score, a written report showing where the leaks are and how serious each one is, and a deep-dive call with me. You leave with a fix list: what to sort before you automate, and where AI will actually help.
Cost: $99, including the deep-dive call. That's considerably less than a month's subscription to most AI platforms.
Frequently asked questions
What is agentic AI?
AI that takes actions rather than just answering questions. A chatbot tells a customer your opening hours. An AI agent books their appointment, sends a reminder, and updates your CRM.
Should I wait before using AI in my business?
No. But you should fix the journey first. For most SMEs that's a matter of weeks, not months, and it makes whatever you automate afterwards far more effective.
Can AI fix a slow response time?
It can make the first reply instant. It can't make sure someone follows up, keeps promises or looks after the customer after they've bought. That still needs a process and a person who owns it.
Do you sell AI tools?
No. YOURCXC is independent. I help you work out what to fix and what to automate, so whatever tool you choose is set up to work.
Find your leaks before you automate them
AI will make your customer journey faster. Make sure it's making the right things faster.
Want to talk through your AI plans first? Book a free 45-minute Fix-It Call.
Comments