Is AI Really Transforming Customer Service? Or Just Handling the Mess?

Every week brings another headline about AI reinventing customer service.

“AI agents now handle thousands of support requests.”
“Chatbots are deflecting 40% of inbound interactions.”
“Digital assistants are reshaping contact centres.”

All true — and yet, here’s the uncomfortable question:

If AI is transforming customer service, why are customers still reaching out so much?

We’re Still Fixing the Wrong End of the Experience

Let’s be clear: AI is doing a great job at what it’s asked to do; resolve tickets, respond to queries, and automate responses at scale.

But those queries and tickets often shouldn’t have existed in the first place.

Customers rarely contact support just for fun. They do it when something is missing, confusing, delayed, broken, or unclear.

That’s not a service issue. That’s a design issue. A strategy issue. A gap between what the customer expected and what they got.

Most avoidable interactions aren’t a measure of demand, they’re a symptom of friction.

We’re optimising the end of a broken journey, not fixing the journey itself.

Brian Solis (Global Innovation Evangelist at Salesforce) put it simply:
“Digital transformation fails when it prioritises technology over people.”

The AI Flywheel: A Smarter Loop for AI and Experience Strategy – excuse the crudity of my drawing, I did try to get AI to tidy it up with little success.

To shift from firefighting to foresight, we need to stop asking “How do we automate more?”
And instead ask: “Why are customers contacting us in the first place?”

Here’s a framework I use to reframe AI planning: the AI Flywheel.
It’s an iterative loop that prioritises strategy before automation, and gets more effective over time.

• 🎯 Start with CX Strategy
  What’s the customer trying to achieve? Where are the moments of effort, confusion, or abandonment?
• 🧠 Define AI Strategy & Use Cases
  Where can intelligence support customers or employees? Not just to automate, but to enable.
• 🔎 Shape Your Data Strategy
  AI needs context. That means clean, timely, relevant data, structured around outcomes, not systems.
• 🚀 Deliver the AI Experience
  Now build. But build for measurable improvement in the journey, not just volume reduction.
• ↻ Refine, Expand, Repeat
  Review feedback. Track outcomes. Iterate. AI doesn’t stand still, neither should your strategy.

Why Human-in-the-Loop Still Matters. Especially in Design

As AI tools evolve, there’s a temptation to remove humans altogether. But when we do that too early, we lose the nuance that makes a service experience feel seamless, empathetic, and human.

In AI development, this is called Human-in-the-Loop (HITL) – keeping human input in the training, testing, and refinement process.

In CX, it should mean:

• Involving frontline human agents in AI use case design – they know what customers really struggle with.
• Observing real customer journeys – not just relying on dashboards.
• Building cross-functional design teams – so marketing, product, and service actually talk to each other.
• Creating smart escalation – so customers who need a person aren’t punished with longer wait times or dead-end loops.

HITL isn’t just for AI tuning – it’s a mindset for building experiences that adapt, evolve, and feel real.

Why Do We Keep Defaulting to Reactive AI?

This isn’t just a strategic failure. It’s human nature and organisational inertia.

🧠 Behavioural Traps:

• Streetlight Effect: We optimise what we can see, high case volumes, and ignore upstream friction.
• Fast Thinking (Kahneman): It feels good to solve what’s obvious. But root cause analysis? That takes effort.
• Maslow’s Hammer: If you’ve just bought an AI platform, everything looks like a chatbot opportunity.

🚱 Structural Realities:

• Short-Term ROI Pressure: Chatbots promise savings this quarter. Journey redesign takes time.
• Data Access: Complaints and tickets are structured. Experience friction is harder to measure.
• Silos: Service owns AI delivery. Marketing owns customer insights. IT owns the systems. Nobody owns the journey.
• Legacy Complexity: It’s easier to add a bot than rebuild the checkout flow or claims process.

Reactive AI still adds value but if it’s your whole strategy, you’re solving symptoms, not systems.

The Real Win: Reducing Service Demand, Not Just Handling It Better

The best service interaction is the one that never needs to happen.

Imagine:

• Customers onboard themselves without questions.
• Notifications preempt “Where is it?” messages.
• Confusing forms are simplified or rewritten by AI before errors happen.
• Your agents have time to solve meaningful problems, not reset passwords.

This isn’t some theoretical dream. This is what’s possible when AI, data, and CX strategy are aligned before delivery.

What the Experts Say

You’re not alone if this feels like a shift in thinking. And the best minds in the space are saying the same thing:

• Paul Greenberg: CRM isn’t just a system, it’s a culture. Automating friction doesn’t build loyalty.
• Josh Bersin: AI isn’t just a tool, it demands a rethink of how work is done and who it’s for.
• Forrester: Organisations that use AI for proactive personalisation see longer-term gains than those focusing on call deflection alone.
• McKinsey: Up to 70% of AI’s potential value in service comes from enhancing customer journeys, not just reducing costs.

Where to Start Instead

If this resonates, here are five questions to kickstart a better approach:

1. What are your most common inbound service interactions and which ones shouldn’t exist at all?
2. Where are customers hesitating, dropping off, or escalating?
3. What parts of your service feel designed for the company, not the customer?
4. What would you learn if you asked your agents why customers really call?
5. What happens if your goal isn’t deflection, but elimination of the pain that caused the contact?

What’s Coming Next

In my next post, I’ll explore real-world examples of organisations applying this mindset, some who got it right, others who launched automation too early and paid the price.

Because while AI keeps getting smarter, the real test isn’t technical.

It’s strategic.

And sometimes, the boldest move isn’t building the next bot, it’s asking whether the customer should have needed help in the first place.

Like this post? Disagree entirely? Good. Let’s talk.

I’m always up for a conversation about where AI meets customer experience, especially if you’re wrestling with strategy, design, and the beautiful mess in between.