For contact center leaders trying to work out where AI belongs first, interactive voice response (IVR) is an unusually strong candidate. The friction is high, the cost is easy to measure, and the downside of getting it wrong is limited.
The conversation around AI often jumps straight to predictions about the fully automated AI call center, but that is not how the shift is happening in practice. Many organizations are starting somewhere much more practical: improving one of the most frustrating parts of the customer journey.
Customers know the experience well:
“Press 1 for billing. Press 2 for technical support. Press 3 for all other inquiries.”
If none of those options fit the problem, callers press zero repeatedly, choose the closest option, or start the process over again.
Call center IVR is one of the few systems every customer has an opinion about, and for contact center leaders looking to introduce AI into operations, it is often a logical place to begin.
Why interactive voice response is the natural starting point
There are two reasons IVR stands out as a starting point for AI. First, replacing an IVR system with Voice AI carries less risk than many other applications of AI in a contact center. Second, it addresses an experience customers already want improved.
When Voice AI answers a call in place of a traditional IVR system, its initial responsibility can be relatively contained: understand why the customer is calling and route the call correctly. It is not yet being asked to move money, change account information, make eligibility decisions, or complete other consequential transactions.
Voice AI can do those things too, but every additional responsibility requires deeper integrations, more testing, stronger safeguards, and ultimately more trust in the AI.
That combination—high customer frustration and a relatively contained first use case—makes IVR an unusually attractive place to introduce AI.
Replacing a menu tree does not require a company to redesign its workforce. There is no customer loyalty attached to “Press 4 for more options.” The menu is simply where time in IVR accumulates and where abandonment starts. And unlike many technology initiatives, the improvement is immediately visible to the customer.
Instead of forcing callers to navigate a list of options, the system can simply ask, “How can I help you today?” and determine where the request belongs.
One middle-market fintech company chose IVR replacement as its first Voice AI implementation and is on track to save $1.4 million in its first year. This was not a global enterprise. It was a growing company that needed measurable operational results before expanding further.
The company reduced misrouting and achieved a twofold improvement in calls resolved on the first transfer, creating momentum for additional AI initiatives across the business.
That last point matters. Successful AI adoption is partly a question of trust.
Imagine an employee who handled one million calls with 99.6% accuracy. You would probably trust that employee with more responsibility. AI should earn responsibility in much the same way.
Give it one clearly defined responsibility. Measure how well it performs. Then expand what it is allowed to do.
The organizations that get this right will have the vision to see what AI can eventually become and the discipline to earn that future one responsibility at a time.
As Newo Co-Founder David Yang explains:
“I visited a number of call centers in Europe this summer and a common question is, ‘Where do we get started?’ I like to compare it to hiring and training a new employee. Before you post the job, you write the job description. When you map the risks and return on investment, IVR consistently stands out as the application that should be implemented first.”
The economics are compelling
A typical four-minute interaction can cost roughly $2.75 when handled by an onshore agent and around $1 when handled offshore. An AI voice agent can often handle a similar interaction for approximately $0.24.
But focusing only on labor rates misses another important part of the equation: attrition.
High turnover remains one of the most expensive challenges in contact center operations. The real cost of labor includes salary, recruitment, training, supervision, facilities, and management overhead.
AI can change that equation not only by reducing cost per contact, but also by reducing some of the infrastructure required to support a high-turnover operation.
Understanding a request is not the same as solving it
Introducing AI into customer service is not without challenges.
Imagine a customer says, “I want to check my account balance.”
The AI may understand the request perfectly. But to actually provide the balance, it needs access to the correct account, reliable identity verification, permission to retrieve the information, and a connection to the system holding it.
The same challenge applies to booking appointments, changing addresses, tracking orders, making payments, or resolving billing issues.
Understanding what a customer wants is only the beginning. Completing the task is where the real value exists.
Routing is a useful first responsibility because the consequences of an error are relatively contained. Once an AI agent begins changing account information, processing payments, scheduling appointments, or taking action inside business systems, the operational requirements become much higher.
This is where AI moves from understanding conversations to actually doing work.
Integration becomes critical. An AI agent has to work inside the systems a business already uses, including systems that may not have modern APIs. Newo addresses this by operating those systems the way a person would, allowing the agent to move beyond understanding a request and actually complete the work.
A more practical path to the AI agent
For most contact centers, AI adoption happens gradually:
- Replace the menu with natural language.
- Automate repetitive work.
- Get the human handoff right.
Step three is where IVR replacement can fail quietly, so it is worth being specific about what “right” means.
A blind transfer is not a handoff.
If the caller must provide their name, account number, and issue a second time, the organization has simply moved the friction instead of eliminating it.
A Newo AI agent can transfer the conversation context, including what the caller wanted and what has already been verified or completed. The representative can begin solving the problem instead of restarting the conversation, which takes real time out of AHT.
Preserving context can also improve First Contact Resolution (FCR) by reducing unnecessary transfers and preventing information from being lost as a call moves through the organization. Fewer transfers and fewer repeat contacts ease queue pressure, which shows up as lower ASA.
At CCW 2026, David Yang highlighted an important distinction in how Newo approaches Voice AI:
“We didn’t talk about eliminating jobs with AI. We talked about how Voice AI can handle repetitive tasks so team members can focus on the most valuable calls. We are consistently improving First Contact Resolution rates by replacing IVR systems with a ready-to-deploy Voice AI agent.”
Building an effective Voice AI solution involves much more than connecting a knowledge base to a voice model. Production deployments require integrations, permissions, testing, monitoring, escalation logic, and a clear understanding of what the AI is and is not allowed to do.
David Yang compares AI onboarding to training a new contact center employee. You would not hand a new hire a headset on their first day and expect them to manage billing, scheduling, escalations, payments, and multilingual support all at once.
You give them a defined responsibility, measure their performance, and expand what you trust them to do over time.
The same principle applies to AI.
An agent that successfully routes calls may eventually qualify callers, schedule appointments, update records, collect payments, or resolve entire categories of requests. Each added responsibility raises containment.
The goal is not to limit what AI can do. It is to create a practical path toward everything it eventually could do.
The future is earned one responsibility at a time
The AI call center is not going to arrive in one enormous leap. It will happen through a series of responsibilities handed to AI as organizations prove that it can perform them reliably.
Some interactions will increasingly be handled end to end by AI. Others will continue to require human judgment, empathy, negotiation, or decision-making.
The important question is not simply, “What can AI do?”
It is: What have we proven we can trust it to do next?
Start with the IVR. Get routing right. Build trust. Then give the AI more to do.
The vision can be ambitious. The implementation should be earned one responsibility at a time.
The first call you should make
Replacing interactive voice response is the lowest-risk place to begin. Newo’s Creator can read your website, understand what your business does, and create a working AI agent you can call and test yourself.
Build your own Voice AI agent and experience the difference for yourself.
