Senior AI & UX Specialist Itai Bercovier breaks down why "set it and forget it" AI strategies are quietly costing companies their customers and what separates the businesses getting it right in the world of AI in customer service.
“AI offers huge potential in making businesses money and keeping them ahead of the curve,” Itai opens. “But whilst the potential is great, the strategy for effective AI in CX is delicate. Following everyone else and handing everything over to tech is not the winning formula. You’ve got to stay in the driving seat.”
AI myths
Two thirds of UK contact centres are already using AI (MaxContact 2025/2026), but Itai warns that AI is not a ready-made solution. “There’s a lot of work in fine-tuning AI, which is often compounded by a company’s lack of expertise. AI is a new paradigm. Even if companies have great IT departments, they probably don’t have a dedicated AI expert embedded into the business and its decision-making.” The impact of this missing piece is huge. Just take the UK retail sector where firms have lost up to £8.6bn due to negative customer AI experiences (CEBR/Trustpilot, November 2025).
Training is vital
In order to support various roles within one business, AI needs teaching about that business, Itai insists. “Take appointment booking for example: this will look different for a restaurant, versus a salon or estate agents. AI needs context and tuition, just like staff do, to become good at its job.”
AI ≠ downscaling
"The beauty of AI isnt in reducing your human workforce, its upskilling it. Your help centre should have time to deal with the complex queries that can’t be handled by AI, whilst AI relieves them of the simpler queries."
Popular ≠ best
Emotional nuance
“Tech departments and CX leads are sold the idea that AI can use your brand voice, recognise emotions, and respond logically with the right action – but it needs to go further“ Itai says. “AI can’t just sound like your brand, it needs to understand your customer’s mood and intent in order to give them the best service, journey and outcome. AI also needs to understand when to be creative and playful, and when to hand over to a human. Say you have a situation where an account holder has passed away, without recognition training AI might regurgitate company policy or flatly say ‘I’m sorry’. Basically, how do we avoid annoying people and give them a really good service? As a customer, I would be annoyed if an AI told me it was sorry for my loss and tried to mimic emotional understanding, because I know it’s just zeros and ones.”
Connection with customers
“Building rapport comes down to things you can’t really put your finger on, like emotion and trust. For tech leads to get the best out of AI in CX, they need to create rapport with their customers,” Itai shares. “Rapport creates trust, which creates customer loyalty. Customer interactions are where rapport and trust are built, which means negative CX experiences can quickly damage customer loyalty. Interactions between your AI agent and human help centre should be seamless for a customer.”
Why AI is failing customers
“Fragmented product ownership,” says Itai. “If your IVR, e-mail and chat channels all have different owners, you end up with separate products with different capabilities. Let’s say I want to complain about something I bought. In a chat, I can upload a picture of the issue, but on a voice channel maybe I need to give the product number. On the phone, an agent would give a maximum three instructions. An AI agent needs to be able to handle all channels, tailoring to each of their strengths whilst giving the same outcome. That doesn’t work out-of-the-box.”
Essential factors when implementing AI in CX
“Technology, brand voice and AI behaviours are all elements that make up the customer experience with AI. Take these into consideration on day one and will save a lot of questions and troubleshooting later.”
Avoiding hidden costs down the line
“AI is seen as a quick and easy fix, but done poorly it can lead to expensive reworks later on,” Itai reveals. “The chances of succeeding are much higher if you take a staggered approach. Don’t try to add your entire knowledge base right away for example. Start with just one intent, test it, find the issues, fix what you need to fix, then you can expand. That’s not to say you can’t fix things post-launch once your AI is already operational; that’s always possible.”
The indicator it’s working
“A customer journey should bring joy,” Itai smiles. “Even if there’s an issue someone’s not happy about, if the problem was solved and the interaction is positive, then it will bring joy overall. That’s the goal.”
£8.6bn says not everyone is getting it right.
What are you doing about yours?