The numbers on enterprise AI projects are hard to ignore, and harder to justify. MIT’s Project NANDA found that 95% of organizations see no measurable return from their AI investments. BCG puts the number of organizations creating substantial AI value at just 4%. CX leaders admit they’re struggling with the gap between AI investment and returns. In the Genesys 2026 State of Customer Experience Report, 42% of CX leaders surveyed cite demonstrating AI value or ROI as one of their top technology adoption challenges, even though 27% say proving ROI is a strategic priority.

None of this means AI can’t deliver value. The problem lies in how organizations plan and deploy their AI initiatives.

At Xperience 2026, Genesys’ Juan Perez, Director of AI Solutions & Services, and Niels Kusters, Senior CX Advisory Consultant, laid out a practical, five-step method for building AI use cases that deliver measurable value.

 

Value Lives in the Friction 

Most organizations start AI initiatives by asking the wrong question: What can this technology do for us? That sends them straight into a catalog of capabilities, like virtual agents, routing, predictive engagement, and copilots, which leads to a search for somewhere to apply each one. Focusing on technology first feels logical. But it leaves organizations with a portfolio of technically impressive AI features that don’t deliver outcomes that matter to the business.

The better question is: Where are your customers struggling? Where do they abandon a task, escalate in frustration, or call back three times to get the same issue resolved? That friction is expensive, as the numbers Perez and Kusters unpacked at Xperience show: 84% of retail shoppers say they’d consider switching brands after a poor returns experience; 67% of telecom customers would leave after unexpected charges. Friction doesn’t just annoy customers. It costs organizations their loyalty.

Starting from friction rather than capabilities keeps you anchored to business outcomes instead of technical possibilities.

 

Seeing the Whole Iceberg

Perez and Kusters used an iceberg visual to illustrate how easy it is to start with the wrong focus. At the very top of the iceberg are the AI capabilities, which are what typically comes to mind first as organizations contemplate deploying AI. Just below the AI models is the technology and data layer: the platforms, integrations, and data pipelines that make the AI models usable.

Most of the iceberg sits fully submerged and out of view. It includes the people, processes, and change management required for successful outcomes. This is the part almost nobody budgets for. It’s also the part that sinks most AI initiatives. While critical, it’s not enough to have the right capabilities and data pipeline. You need employees who trust the AI tools enough to use them, supervisors who’ve bought into a new way of measuring performance, and workflows redesigned around what AI changes in your CX operation.

Skip that layer, and you get a technically functional deployment with low adoption and even lower returns.

 

Five Steps to Build AI Use Cases that Deliver Value

Here are the five steps for building your AI use case that will help you work through the entire iceberg to realize compounding returns on your AI investment.

Step 1: Start with Experiences that Matter

Before considering solutions or AI capabilities, identify which customer experiences make the biggest impact on your business, such as onboarding, billing, claims, returns, and scheduling. These are the journeys where a good or bad experience visibly makes or breaks loyalty and affects revenue. Those impacts tie the experience to a business case with a specific metric, like containment rate or retention, that proves the use case’s value. This step keeps the entire effort tethered to business impact rather than technical innovation.

Perez and Kusters grounded this with a real example: a telecom operator handling more than 1.5 million voice interactions a month identified billing as one of its highest-stakes experiences —the kind of journey where a single bad interaction can undo months of loyalty-building.

Step 2: Find the Friction

Map the experiences you’ve identified, and pinpoint where things break down. Walk the journey the way a customer does: the confusing bill, the bot that loops the same generic FAQ, the repeated explanation to another agent, the resolution that leaves the customer unsure whether they’re still on the hook for a late fee. Each moment of friction is an opportunity for intervention — and adding value.

Forcing customers to repeat information is one of the clearest signals of broken orchestration. Genesys’ 2026 State of Customer Experience report found that 95% of consumers want their information carried across channels so they don’t have to repeat themselves, yet 52% of companies still don’t pass it along to a human agent.

Mapped out, the telecom billing journey looked like this: a customer gets an unexpected charge, turns to self-service and gets looped through generic articles, calls in and has to explain the issue from scratch, gets bounced between teams and starts over each time, and finally gets a ticket with no clarity on the late fee. Five moments, and at least three of them are pure friction.

Step 3: Practice the Art of the Possible

Once you’ve found the friction, resist the urge to jump straight to a solution. Perez and Kusters recommend a classic technique to look at the problem through three lenses. The first is unconstrained — no budget, no legacy system, no limits. At each point of friction or moment of truth in the customer journey, what’s the best-case scenario? Then take a second, deliberately critical pass: Where are the real risks, security concerns, and technical limitations? The third pass takes a balanced approach, optimistic but grounded. These three passes keep you from defaulting to whatever your team assumes is most feasible.

Running the billing journey through all three lenses produced a redesigned version: the customer gets a proactive alert before the charge feels like a surprise, a guided self-serve flow resolves the simple cases, and when a live agent is needed, they already have full context and can close the issue in one contact.

Step 4: Design a Solution Around Your Vision

With a clear picture of what’s possible, get specific about how you’ll build the solution. Identify the technology and platform capabilities you’ll need, what data must be in place and what metrics will tell you whether the use case is working. Use case KPIs often differ from the business metrics they ultimately serve. You’ll need both to know it’s working.

For the billing journey, the vision translated into specific use cases with their own KPIs: an agentic self-service journey to lift containment rate, predictive routing to improve first contact resolution, and Genesys Cloud Agent Copilot to give agents real-time guidance to improve the customer experience.

Step 5: Build, Pilot and Monitor

It’s crucial to ensure that you’ve defined a use case that will provoke positive changes and deliver value that compounds over time. So, before committing resources, run the use case through a readiness check across four dimensions:

  • Does it clearly serve the business case?
  • Is it technically feasible?
  • Is the data in place?
  • Is the organization ready for the change?

Then launch small, measure constantly, and treat the deployed AI as something that needs ongoing feedback and retraining, not a project you finish and walk away from. An AI system without a feedback loop drifts out of sync with the outcomes it’s supposed to deliver. Measuring KPIs in real time, analyzing signals daily, identifying drift weekly, retraining monthly, and validating quarterly is what keeps an AI system improving instead of decaying. Ensure that KPIs track both the technical solution and the business outcome it’s meant to deliver.

 

Ripples and Compounding Effects

One of the biggest missteps with AI initiatives is overlooking how far a single automation decision can reshape the rest of the operation. Introduce one capability at the core of a workflow, and it sends ripples through routing logic, agent tooling, data governance, even how you measure success.

The instinct is to treat this as risk to manage. Treat it instead as a signal that you’re doing things right. Perez and Kusters offer strong advice: don’t just tolerate these ripples; actively provoke them. They’re the mechanism behind compounding returns.

Provoking ripples only drives those returns if you’re starting with the right use case. That’s why the readiness check in step 5 above is so important. Your use case must serve the business case, be technically feasible, depend on data you have in place, and be supported by agents, supervisors and process owners.

The more use cases you roll out, the more impact they make because the data, trust, and infrastructure from earlier use cases carry forward. A single well-placed use case can deliver value on its own. But value compounds when AI stops being a point solution and becomes foundational to your CX.

 

Adapt, Lead and Scale

How you scale matters as much as where you start, so ramp up deliberately. Prove the concept with a small pilot, stress-test it, expand gradually, and only move to full production once the feedback loops and safety checks are in place. It’s a more patient path than the all-at-once rollouts that used to define contact center technology, but AI behaves differently than the systems that came before it. With AI, scale is something you earn in stages, not something you achieve with the flip of a switch.

Scaling value requires discipline, just like the five-step methodology. So, start with an experience where your customers feel the friction, tie it to a business case your leadership cares about, and let each use case become the proving ground for the next one. That’s the difference between chasing capabilities and building toward agentic orchestration, where people, AI, data and tools work as a unified system. The organizations that are already realizing substantial value didn’t get there by just deploying more AI. They align their AI initiatives — and their people, processes and change management — to the outcomes that matter.