Blog by Josef Novak, Co-founder and Chief Innovation Officer at Spitch AG

Generative AI is changing customer service, but not in the simplistic way many market narratives suggest. It does not magically solve omnichannel communication challenges. What it does is make contextual, channel-aware interactions far more practical – provided the enterprise architecture is ready for it.

That distinction matters. As Gartner’s recent analysis on agentic AI rightly notes, the market is moving faster than the supporting infrastructure, and many organizations are still using agents mainly to automate existing workflows rather than redesign them. For CIOs, the value of AI in customer experience will come less from autonomy for its own sake and more from disciplined integration, governance, and human oversight.

Continuity is the real customer expectation

Customers do not think in terms of channels. They expect one coherent conversation, whether they move from phone to chat, from app to voice bot, or from an agent to self-service. The technical challenge is not just understanding what the customer said but preserving the thread of the dialogue across modalities.

This requires three things:

1. a unified identity layer that connects phone numbers, accounts, app identities, and other signals into one customer view;
2. contextual relevance assessment, so the system can determine whether a new interaction is a continuation of an earlier one; and
3. dynamic history handling, so AI accesses the right mix of long-term history, recent turns, and current intent without overloading the model or exposing unnecessary data.

In practice, context management becomes a design discipline. A well-built system does not simply remember more; it remembers what matters, when it matters, and in the form best suited to the channel.

Governance is the enabler, not the afterthought

Too many legacy automation tools are being relabeled as AI agents. For CIOs, the answer is not to buy into the hype cycle, but to ask harder questions: What exactly is the agent allowed to do? Which model is used? What data is sent to it? Where is inference performed? What is logged, retained, and reviewed?

These questions are especially important in regulated environments such as Switzerland. Data sovereignty is no longer a blocker to innovation – many models can now be deployed locally or within the EU – but it must be built into the architecture from the start. Governance needs to cover model access, retention, fine-tuning policies, consent, and auditability. AI readiness is as much a governance challenge as a technology one.

Local language matters too

In the Swiss market, context awareness also means linguistic awareness. Voice systems must understand Swiss German dialects accurately, while chat responses usually need to be delivered in standard German. That is not a minor localization detail; it is part of the customer experience. A context-aware journey must adapt to how customers actually speak and write, not force them into a one-size-fits-all model.

Multimodality changes the architecture, not the objective

Generative AI is also changing how customer interactions are delivered. A modern model can present structured content, links, explanations, or visual references in chat and then translate that logic into voice. In the same way, it can convert a voice interaction into a written thread with explicit confirmations.

This matters because modality switching is one of the biggest sources of friction in customer service. A CIO who wants a truly seamless journey must ensure the underlying platform can carry context across modalities, not simply across sessions.

Human-in-the-loop is still the right operating model

In customer service, poorly governed AI creates cascading failure risk. The safest and most effective design is collaborative AI: systems that augment human agents with better information, better summaries, and better recommendations.

That also improves employee experience. Agents are not being replaced; they are being equipped to handle complexity more effectively. For CIOs, that is a better adoption model because it reduces operational risk while improving productivity.

Measuring success: beyond cost reduction

If AI is measured only by cost savings, it will be used too narrowly. CIOs should look at broader metrics that capture operational and business impact.

AHT still matters, but it is no longer just a simple duration metric. It now reflects the full cost of an interaction, including model choice, token usage, scenario design, escalation strategy, and human intervention. A large model may be justified for a complex, high-value case, but unnecessary for a simple FAQ or routing task.

The goal remains the same: minimize cost while maximizing customer satisfaction. But cost is now multi-dimensional, and the right metric depends on the use case.

There is also growing interest in turning the contact center into a revenue center. AI-driven sales insights, next-best-action recommendations, agent hints, and cross-center campaigns can help generate revenue as well as improve service.

As AI agents gain more autonomy, additional metrics become relevant: tool and task completion rates, escalation quality, and Customer Journey Improvement trends. Success is no longer just about cost reduction, but also efficiency, quality, revenue impact, and improved customer and agent experience.

Three takeaways for CIOs

• Prioritize continuity over channel automation: The real challenge is preserving context across the customer journey, not simply deploying AI in each channel.
• Design for semiautonomous AI with human control: The collaborative AI enterprise model keeps humans accountable for consequential actions while AI supports speed and consistency.
• Build governance into the architecture from day one: Data sovereignty, auditability, model access, and fine-tuning policies must be defined before scaling AI.

For CIOs, the winning strategy is clear: build a customer service architecture that supports semiautonomous AI, preserves continuity across channels, respects local compliance requirements, and keeps humans in control of meaningful decisions. That is how agentic AI becomes enterprise value rather than enterprise risk.

Contact us to learn how to design a context-aware customer journey that combines generative AI, governance, and human oversight in one scalable architecture.