38 A Smarter City
AI is becoming your front door. Is it open to everyone?
Deen Dayal Yadav explains why customer-facing AI ' s greatest risk is not failure, but a customer who quietly gives up
Most failed AI conversations leave no complaint behind. A customer repeats their postcode twice, corrects their surname for the third time and hangs up. Someone else receives an irrelevant chatbot answer and closes the page. Both go elsewhere. The business may never know why.
AI is moving from the back office to the front door. It answers calls, books appointments, handles enquiries and takes payments. That can make a business faster and easier to reach. It also puts technology in charge of an important first impression.
London is an excellent place to test whether that promise is real. Its businesses serve customers with different accents, languages, ages, access needs and levels of digital confidence. One may be calling from a noisy van. Another may need longer to explain a request. Someone may switch between languages or use a name the system repeatedly mishears.
The testing gap
AI demonstrations usually happen in calm, predictable conditions. The speaker knows what to say. The request fits the script. Nobody interrupts, changes their mind or asks an awkward question. Real conversations are rarely that convenient.
A common and costly mistake is to let the people who built the system do most of the testing. They know the route through it. Customers do not. Frontline staff often provide the best test cases. They know which names, addresses and requests are regularly
misunderstood. Those everyday exceptions reveal more than another polished supplier demonstration.
Invisible failure
The danger is not always that the system crashes. It is that it appears to work.
A dashboard may show that every call was answered, but not that a customer repeated the same detail four times before giving up. It may count chats started without showing whether anyone solved their problem. Activity looks healthy while opportunities quietly disappear.
This is more than an inclusion issue. It is responsible AI in practice. AI should widen access, not quietly narrow it. Friction at the first contact point means lost enquiries, additional support work and weaker trust. If particular groups of customers struggle repeatedly, a business may be narrowing its market without realising it.
Test the messy reality
Before launch, give a varied group of real customers tasks to complete without telling them which words to use. Include different accents, devices, names and postcodes. Add background noise, interrupted sentences and unexpected questions. Watch where people hesitate, repeat themselves or look for a way out. Those moments are more valuable than a flawless demonstration.
Testing should continue after launch. Customer habits shift, services change and systems evolve. A conversation that worked last month may fail after a small change elsewhere. Review samples of real conversations, remove personal details and listen for the moment when confidence begins to disappear.
Measure the journey
Once the system is live, measure outcomes rather than activity. An answered call is not necessarily a completed enquiry. A fluent conversation is not necessarily accurate. Review abandoned calls, failed bookings, corrections and requests for human help. Most importantly, check whether the handover works.
Customers should always be able to ask for a person without navigating a maze. Human handover is not a defeat for automation. It is part of good service.
Smaller businesses may have an advantage here. Their staff often know the real exceptions: the customer who is unsure what to ask for, the family member calling on someone else ' s behalf, or the enquiry that does not fit a standard form. That knowledge should shape the system.
The real test of customer-facing AI is not how impressive it sounds. It is whether a customer can get something done without feeling ignored, misunderstood or trapped.
AI can keep the front door open around the clock. But a front door only works if people can get through it.
Deen Dayal Yadav is founder and director of Softomate Solutions Limited
www. softomatesolutions. com