The AI Dilemma: Why the Promise of Better CX Falls Short

When it comes to artificial intelligence (AI) in customer experiences (CX), too many leaders are asking the wrong question. The debate shouldn’t be “How fast can we scale AI?” But rather, “Is our system ready for AI at all?”

Because here’s the uncomfortable truth: in regulated industries—finance, healthcare, government—fast AI deployed on shaky foundations doesn’t enhance customer experience. It erodes consumer confidence.

AI Acceleration and the Gap It’s Creating

The deployment of artificial intelligence has exploded. According to National University, 74% of businesses plan to increase their use of AI-powered chatbots. Yet Gartner reports that 64% of customers would prefer companies not use AI for customer service at all.

This isn’t tech resistance. It’s a system breakdown. Customers are interacting with AI that hasn’t been designed with transparency, oversight or escalation in mind. And when AI fails in high-stakes moments—disputes, denials, emergencies—it’s not just a poor experience. It’s a breach of trust.

The Real Risk: Automation Without Augmentation

In emotionally charged “life moments,” automation often isn’t just ineffective, it’s damaging, or even dangerous. Imagine disputing a fraudulent credit-card charge or navigating a denied medical claim for vital treatment, and the only “help” is a chatbot with no authority and no empathy.

That’s the breaking point.

Artificial intelligence is exceptional at pattern recognition, but it has no intuition for emotional context. And it certainly can’t build rapport in a crisis. We’ve seen this play out across sectors: when automation is used as a deflection tool, rather than a force multiplier for human support, it breaks the experience and the relationship.

That’s why leading customer-experience organizations don’t ask “How much can we automate?” They ask: “Where must we protect the customer?”

It’s Not the AI. It’s the Architecture.

One of the most common misconceptions I see is blaming the service model when the real issue is the system around it.

In Working Solutions’ case, we integrate AI into our CX systems to ensure it’s part of a larger orchestration, not the driver. That means:

• Clean, structured data foundations.
• Clear escalation paths when AI can’t resolve an issue.
• Role clarity between human agents and AI assistants.
• Compliance reviews baked into design, not added as patchwork.
• Outcome-based quality assurance (QA), not just compliance sampling.

You can’t just layer artificial intelligence on top of dysfunction and expect it to work. You’ll only scale the dysfunction.

A recent Harvard Business Review article found that after Autodesk introduced transparency “AI cards” for customers, their trust and reliability ratings jumped significantly. Trust isn’t a side effect; it’s a system feature. But only if you design for it.

Getting the Blend Right: Where Humans Still Matter Most

AI shines in structured, low-stakes scenarios: “Where’s my order?” “What’s my balance?” “What time is my appointment?”

But in regulated industries, many moments carry high emotional weight—and legal risk. Those aren’t just “support tickets.” They’re make-or-break moments for loyalty and brand equity.

That’s where blended systems win. The goal isn’t to replace labor with AI. It’s to make human labor more impactful. Working Solutions calls it: preserving confidence at scale.

When humans are supported by AI—but still lead in high-context, high-stakes situations—we see stronger resolution rates and improved customer satisfaction (CSAT). And most importantly, we protect the trust integral to the client brands we represent.

From Compliance to Intent: The QA Evolution

Legacy QA systems sample 10–15% of interactions. But with AI-enabled tools, we can assess 100%. That’s not just for script adherence, but for intent alignment and outcome quality.

This shift, from compliance auditing to intent auditing, is one of the most important undercurrents in CX today. The question isn’t “Did the agent say the right phrase?” It’s “Did the system, human or AI, solve the right problem in the right way?”

And when something goes wrong, we’re not just looking for who to blame. We’re asking how to refine the system to be more efficient, build equity and customer confidence.

Smaller Models, Smarter CX

There’s a lot of talk around the benefits of large language models (LLMs). But in regulated industries, broad isn’t necessarily better.

At Working Solutions, we’ve begun experimenting with small language models (SLMs) based on 30 years of proprietary data in contact center outsourcing. These focused, secure systems have achieved more than 90% accuracy on scheduling and forecasting, without the irrelevant or inconsistent data typical of large, internet-trained models.

SLMs enables us to optimize for select context, privacy and compliance. They don’t randomly pull information from TikTok. They only know what they’re supposed to know, within well-prescribed parameters we set.

This is the future for CX in complex environments: less broad application, more surgical precision for the exact data to deliver specific results.

What Really Scales: Trust

The truth is, anyone can scale AI. But not everyone can scale trust.

Customers won’t remember how fast your chatbot responded. They’ll remember whether it helped, or whether it made them feel assured when they needed support most.

The metrics that matter are becoming more nuanced. While still valid, it’s no longer just average handle time. It’s trust equity: how much confidence your business earns, interaction by interaction.

The winners in AI-powered CX won’t be those that move the fastest. They’ll be the ones who moved most deliberately, with the right architecture, the blend of resources and respect for what humans still do best, which is critical thinking and relation-building.

So, what’s your next step? Take a hard look at the quality of your customer service. Ask yourself: “Are we building trust, or just scaling dysfunction?”