Your customers don't want an AI friend. They want an answer.

Your customers don't want an AI friend. They want an answer.

Our research found 98% of organizations are training their AI to act more like a human. They’re adding things like:

• Typing animations

• Synthetic pauses

• Warmer phrasing

• On-brand scripted personalities

It's close to unanimous across the 660 business leaders we surveyed, and it's a lot of engineering hours pointed at one outcome.

Then, you ask customers what they want from customer support…well, personality is not on the top of their priority list. It’s not near the bottom either. It’s not on the list at all. And while you can make the Henry Ford claim that if you listened, your customers would just want faster horses—it doesn’t quite hold up here.

You don’t have to guess what your customers want, though. They told us very explicitly, and most of it's easier to deliver than human-sounding robots.

What customers want from customer support

Consumers were asked what matters most when they contact customer support. The answers are unglamorous:

• Immediate issue resolution: 40%

• Speed of assistance: 40%

• Accuracy of information: 38%

• 24/7 availability: 38%

Those are all operational attributes. There’s nothing about tone, warmth, empathy, or any of that.

Now, let’s be fair to the 98% because it isn't that customers reject personality outright. 54% do say they want AI with more human-like personalities. But brands put that number at 73%, and nearly half of consumers are neutral or indifferent as long as the thing works. 

Still, that’s a priority disconnect, and it also doesn’t explain why businesses are making humanness a non-negotiable here.

What brands are building for customer support

We know what the customers want, and here’s what the brands are prioritizing:

• Training AI to recognize and respond to customer emotions: 41%

• Personalization based on user data and preferences: 40%

• Natural conversation flows that mirror human interaction: 39%

• AI personalities aligned to brand values: 39%

You see the problem here? None of those implementations are fixing issue resolution, speed, accuracy, or availability. Well, maybe personalization—because knowing your customers could accelerate support and make recommendations more accurate.

But the other three? Not so much.

What does “human” mean to a customer?

Consumers were asked what makes an interaction with a brand feel genuinely human or empathetic. Here's what they said:

• Acknowledging my specific situation or problem: 34%

• Offering multiple solutions when the first doesn't work: 32%

• Using conversational language instead of scripted responses: 30%

• Offering compensation or refunds when something goes wrong: 26%

Almost all of those require personalization. 91% of consumers say personalized interactions feel more authentic. However, they’re not saying personable. They’re saying personalization.

Those aren’t the same thing.

Ultimately, an AI agent can't feel empathy. It can produce interactions customers describe as empathetic, but the mechanism is context. Your customer isn’t feeling the warm and fuzzies because the robot’s voice sounded more understanding. 

No, a customer who doesn't have to repeat their issue feels understood. A customer offered a second option when the first fails feels taken seriously. A customer who moves from bot to human without re-explaining their issue feels looked after.

None of that is personality. All of it is data arriving on time.

Why the theatrics backfire

There's a cost to warmth that doesn't land.

An AI that opens with "I completely understand how frustrating that must be" and then asks for an order number the customer already provided has made things worse than a blunt agent would have. 

The empathy raised the expectation. The failure that followed is measured against it.

39% of businesses already name weak emotional understanding as a top driver of AI dissatisfaction, second only to lost context between channels. Really, most of what reads as a lack of empathy is a lack of information.

Twilio’s Inside the Conversational AI Revolution report found only 39% of consumers describe AI agents as helpful while 51% say they feel robotic. Robotic isn't a criticism of tone from people who received plenty of scripted warmth. It's what a conversation feels like when the system doesn't know anything about you.

Where to build your conversational AI instead

More than half of organizations now optimize for functional effectiveness before personality.

• Give the agent history before you give it a voice. Twilio Conversation Memory maintains persistent customer profiles across sessions and channels, so an agent opens knowing who it's talking to.

• Make context survive the channel change. Twilio Conversations models chat, messaging, and voice as one lifecycle, which removes the repetition customers cite most.

• Design the second option. Offering an alternative when the first path fails ranked second in what customers call empathetic. That's a flow decision. Tone has nothing to do with it.

• Escalate before the warmth runs out. Twilio Conversation Orchestrator triggers handoffs on conversation rules, and the payload into Twilio Flex carries history and a summary so the human doesn't restart.

Again, more than half of your customers want the human-sounding personality. They just want the practicality first.

Useful beats charming (every. single. time.)

Customers experience warmth as being known. Being known requires the system to have their history and surface it in time. Get that right and the tone takes care of itself because an agent that already knows the problem doesn't need to perform sympathy about it.

Want the research behind these numbers? Read the Customer Insights Series guide on why customers value empathy over imitation, or see how context and trust interact in How to Scale Automation Without Losing Customer Confidence. Ready to build? Start with Twilio for free.

Frequently asked questions

Do customers want AI to sound human? 

Yes, but it’s less than brands assume. 73% of brands believe customers want more human-like AI personalities, while 54% of consumers agree. Consumers ranked resolution, speed, accuracy, and availability above personality.

What makes an AI interaction feel empathetic? 

Twilio research ranks acknowledging the customer's specific situation first at 34%, then offering multiple solutions (32%), conversational rather than scripted language (30%), and compensation when something goes wrong (26%).

Should I give my AI agent a personality? 

Yes, by all means, but prioritize context first. 98% of organizations are training AI to act more human, but three of the four behaviors customers call empathetic depend on knowing the customer rather than on tone.

Why does my AI sound robotic to customers? 

51% of consumers describe AI agents as robotic while only 39% call them helpful. The complaint usually reflects an agent operating without customer history instead of the wrong tone.