‘Customers don’t hate AI – they just hate bad AI.’ How to deliver AI that customers love

AI has fantastic potential to transform customer journeys – resolving issues faster, creating seamless interactions and freeing up human agents for the hard stuff. But this only happens when AI is done well.  So, what does good AI look like – and how can you deliver it? 

It’s not hard to find scenarios where customers are left seething with frustration after being pushed toward self-service options. Picture this: a query not covered in the FAQs is met with a chatbot that loops users right back to those same unhelpful FAQs. An older customer, struggling to navigate complex automated menus, just wants to speak to a human — but is told to send a WhatsApp message and wait for a response. Or consider someone trying to track down a missing order, only to be bounced from bot to bot without ever reaching a live agent who can actually help. Chances are, you’ve experienced something similar yourself. It’s no wonder so many people find AI-powered service solutions not just unhelpful but infuriating.  

Poor service drives customers away 

No-one wants customers to feel like this after they’ve reached out for help. But too often, that’s exactly how self-service solutions make them feel. Customers are fed up with bots that don’t listen, don’t help and don’t escalate when needed. In our in-depth survey to explore the real-life service experiences of 1,000+ customers, 55% said they have stopped using a company due to poor service. For many, infuriating chatbots and rigid IVRs have become a symbol of their frustration. Even though self-service solutions have moved on in leaps and bounds, many customers are scarred by bad experiences with poorly implemented solutions badged as AI. 

As a result, it’s no surprise that scaling AI across customer service can feel daunting. While the business case is compelling – faster response times, better agent performance and lower costs – the risks are real. Bad AI not only frustrates customers in the short term. It can damage long-term brand loyalty and, when customers walk away, wipe out customer lifetime value.  

Customers are open to AI, if it works 

There is, however, good news. While customer trust in AI is fragile, it is possible to fix it. Of the 1,000+ customers we spoke to for our survey, 40% told us they are happy to use AI as long as it matches the quality of human support.  

The bar for regaining trust is high. That’s because much of what customers have been sold as ‘AI’ in customer service isn’t AI at all. Our analysis shows that up to two-thirds of current implementations rely on basic automation, scripted decision trees or rule-based bots masquerading as AI. To win back trust, organisations must shift away from shallow solutions that drive operational benefits and cost reduction at the expense of customer experience. What they need to implement is high-quality, context-aware AI.  

Where high-quality AI solutions are expertly implemented, the results speak for themselves. Customers report better service, agents become more effective and costs fall. Great experiences delivered by high-quality solutions become points of differentiation and can drive growth. 

Leading companies are already demonstrating what’s possible: 

• FedEx is using PolyAI in its voice channels globally and AI across its chat channel to provide real-time customer support across easy-to-resolve intents such as shipment tracking, shipment changes, payment processing. More complex queries are routed to agents for resolution based on accurate intent detection and routing. 

• Lufthansa is using Cognigy to provide AI service agents, handling 16m+ conversations per year, reducing handling times and seamlessly handing customers over to agents when their intents cannot be resolved.  

• Decathlon has deployed Parloa across voice, chat and messenger channels to empower its agents, identifying 74% of customers by their order number, eliminating 20% of repetitive tasks.  

How to cut through complexity to deliver AI that works 

These examples prove that customers don’t hate AI – they hate badly implemented AI or self-serve solutions that are masquerading as AI. So how do you deliver the ‘good’ kind? The path to effective AI in customer service is not a quick fix. Our experience is that it takes strategic focus, thoughtful design and technical rigour to balance complex considerations around customer experience, technical complexity and business value. Here are some key principles to guide you on your journey: 

Start with executive commitment and clear vision 

Deploying AI at scale across the contact centre is not a technology upgrade – it’s a full-scale transformation. That means it requires senior-level ownership, cross-department buy-in and a clear, shared vision of success. Without this, initiatives risk fragmenting or stalling. 

Select the right vendor – based on real-world performance 

The AI marketplace is crowded with options, all promising exceptional results. But hype doesn’t guarantee outcomes. Our benchmarking helps cut through the noise by evaluating solutions based on real-world performance.  We’ve undertaken market scanning on both vendors and live use cases to identify what works and what doesn’t. We can also make an expert assessment of your current technology stack to identify whether existing platforms can be leveraged or require enhancement.  

Be strategic about use case selection 

Success depends not only on the technology you choose, but on where and how you apply it. There are two options here. The first is to select individual use cases focused on realising value. For one client, a major energy company, we identified annual savings of £5–10 million using this approach.  

The second option is to target end-to-end transformation of the customer experience and reimagine how you service customers. This involves bringing together multiple use cases across customer, agent and back-end operations – and considering your teams and AI as one. Prioritising the right use cases ensures early wins, accelerates adoption and aligns outcomes with both business goals and customer needs. For example, a recent AI strategy roadmap we developed for a major UK telco staged implementation by complexity and value, ensuring efficient delivery while maintaining focus on high-impact outcomes. 

Reimagine the customer journey and build in smart escalation  

AI performs best when it’s integrated into a thoughtfully designed customer journey – ideally one that reimagines the possibilities rather than simply applying AI to execute components of the existing journey. The new journey should include clear rules for escalation to human agents. The key is to understand not just what AI can handle, but what customers are comfortable letting it handle. For example, for one insurer, we found that, counterintuitively, most customers preferred to report a bereavement via an automated interface rather than speaking to a live agent – demonstrating that human touch isn’t always the default preference. 

Modern AI can intelligently determine when to escalate, based on factors like query urgency, interaction history and sentiment. It can also route customers to agents with the right skills – be it to provide empathetic reassurance or fast, technical resolution – ensuring more personalised, effective support. 

Ensure clean, well-tagged data and smart training pipelines 

High-performing AI starts with high-quality data. Yet too many organisations launch AI service on limited, outdated or poorly structured knowledge bases. In our survey, 34% of customers said chat boxes failed to provide useful answers. This is because the underlying models have often been trained on incomplete data.  

While it’s possible to get started with a basic dataset, long-term success depends on cleaning, tagging and enriching your knowledge base. Building robust training pipelines tailored to your business ensures your AI can learn continuously, adapt to customer needs and deliver service that feels truly intelligent. 

The bottom line 

Good AI doesn't happen by accident. It takes a deliberate strategic approach – from leadership buy-in to smart vendor choices, targeted use cases, thoughtful design and quality data. With these elements in place organisations can move beyond frustrating bots and deliver AI-powered service that customers actually prefer.