Empathetic AI: Why your next customer call might feel surprisingly human

Customers aren’t frustrated by AI itself — only by bad AI experiences. In customer experience, the only AI that truly works puts quality before automation and scale, keeps humans in the loop for regulated or sensitive cases, and is built to be auditable from the start.

AI triages while humans decide, and this is where a more empathetic form of AI can come into its own. It can make customers feel heard and understood while helping human agents to manage their workloads — something that was missing from previous technology-driven approaches to contact-center conversations.

With business leaders under pressure to optimize their operations with AI, could this be the answer to maintaining customer trust while leveraging the benefits of AI?

What happens when AI learns to understand how we feel?

As AI evolved from a theoretical concept into a tangible product, it quickly diversified. First came GenAI, which could create text, images and documentation. This was quickly followed by agentic AI with its ability to understand, reason and act autonomously with far less human intervention. Empathetic AI is the next frontier, switching the focus to one of the main criticisms of AI: its lack of empathy. 

An empathetic AI system is trained to recognize human emotions and tailor its responses accordingly. It decodes sentiment through emotional cues such as pitch, speed and tone of voice. Instead of sticking to a generic script, it analyzes prosody — not just what people say but also how they say it — and sentiment, then follows policy to provide an appropriate response, including through de-escalation techniques.

In other words, by being responsive to the nuances of human conversation, it’s far removed from the stilted, robotic interactions often associated with AI customer-service experiences.

Empathy is the goal; policy decides the path

There are many misguided perceptions about how AI is deployed in customer service. It won’t stop customers from speaking to a human agent, but it will apply triage upfront while its human colleagues handle more complex issues. In this sense, it’s best seen as an addition to human service delivery, not a replacement for it.

The process starts with speech-to-text conversion to capture what the customer says. The system then analyzes pitch, pace and tone to detect sentiment. A policy engine flags signals such as vulnerability, complaints or credit issues and routes the interaction to self-service, assistance or human support.

Behind the scenes, the AI agent suggests replies, surfaces knowledge links and logs every step for transparency.

AI triage: The first line of empathy

Empathetic AI acts as the first touchpoint in customer service — working through chatbots or as the initial responder to calls. It understands the customer request and decides whether it should handle the request itself or hand it off to a person. This frees up human agents to focus on more complex cases while AI agents manage routine tasks.

Handoff policies dictate the process. For example, to maintain customer trust, sensitive situations — such as credit issues, vulnerability indicators (distress, bereavement or accessibility) and complaints — are always escalated to a human agent with full transcripts and context. The human agent retains oversight and can override any suggestions made by AI, while customers can request human interaction at any time.

AI advisory: Empathy in action

Empathetic AI can also support human agents on live calls by analyzing tone and emotion in real time, suggesting responses that de-escalate friction and guide the conversation toward a positive outcome — even as tensions rise and it becomes harder to think of appropriate responses. It can also summarize notes from the conversation for compliance and quality control.

Turning empathy into ROI

In principle, empathetic AI sounds promising, but no organization can commit to digital transformation without knowing where the return will come from. The good news is that with empathetic AI, the value quickly becomes tangible.

Here, ROI goes beyond dollars and cents. It is also measured in time saved, long-term efficiency, brand trust and the lifetime customer value gained from happier employees and loyal customers.

Where empathetic AI delivers a measurable impact:

• Resource efficiency: By using AI to handle repetitive tasks and people to handle more sensitive calls, you make the best use of your resources and optimize staffing levels without compromising on service quality.
• Response times: With AI doing much of the heavy lifting, response and resolution times are shorter. Teams can scale instantly to meet spikes in demand, improving retention and customer satisfaction (CSAT) scores.
• Employee experience: Introducing a tool to manage tickets, handle repetitive tasks and help de-escalate conflict improves employee experience and engagement while supporting employee retention. 

Laying the foundation for empathetic AI transformation

All large digital transformation projects need a clear plan — from strategy and implementation to continuous optimization. You need to map out the emotional journey your new empathetic AI system is designed to support, then train your employees to use it and collect feedback to keep improving it.

Start planning with our “Ship in 90 days” checklist:

• Pick three intents: It’s best to choose common customer queries such as password resets, address changes and fee explanations.
• Establish a baseline for CSAT, quality assurance, wait times and escalation accuracy: Measure your starting point to track how empathetic AI improves performance over time.
• Ship agent‑assist: Start with AI supporting your service team, then enable automation on one task once quality testing is complete.
• Monitor performance for two weeks: See how AI agents support customer interactions and publish the results.
• Evaluate after 90 days: Scale successful AI deployments and discontinue those that don’t deliver results.

Following a structured rollout plan is essential, but real success depends on how you manage quality, risk and customer experience.

3 practical tips from NTT DATA to get started

1. Cap automation by quality, not volume: The goal isn’t to automate everything; it’s to automate well. Publish a clear deflection cap and link it back to CSAT or quality-assurance performance. If the quality dips, automatically route customers back to a human agent.
2. Risk holds the failsafe switch: Give the Chief Revenue Officer the authority to stop the production of models. Run quarterly drills and keep model artifacts for at least 13 months.
3. Ban dark patterns in support: Don’t hide human agents behind chat or interactive voice response loops. Make “speak to a person” easy to find. It’s easier to win through great automation than through forced automation.

Transforming with empathy and expertise

At NTT DATA, we see both the whole picture and the tiny details that make it work. Our goal is simple: to make digital transformation stick.

We work extensively with banks and other financial institutions to review how they use technology, and we apply the latest technologies and best practices to boost their resilience, keep customers happy and help them stay ahead of the competition. Let us show you what we can do for your organization.