Pete Hanlon, CTO of Moneypenny. Moneypenny handles outsourced phone calls, live chat and digital comms for thousands of companies globally.
If your AI voice was a person, would you hire them to represent your brand? Would you put them on reception? Would you want them answering your most valuable customers? Would their tone, pace and manner reflect how you want your business to feel?
Most companies can’t answer those questions because they’ve never really made the decision. The voice arrived as part of an AI agent implementation. Someone clicked through a list, chose a voice that sounded reasonable and moved on to the harder integration work. It felt like a technical setting rather than a brand decision.
However, that voice may now speak for your company thousands of times a day, and for some customers, it’ll be their first experience with your business.
Your website went through design reviews, your logo has usage guidelines and your customer-facing teams are trained in how to speak to customers. Meanwhile, your AI voice may have received only approximately 30 seconds of consideration during setup.
This is a brand decision rather than a technical decision, and increasingly, it may be one of the most important brand decisions you make.
Voice: Part Of The Outcome
It’s easy to treat voice as polish and something to refine once the system works, but callers start forming an impression immediately.
Pace, warmth, tone and accent all affect how a conversation feels. They can influence whether a caller trusts the interaction, how comfortable they’re sharing information, whether they ask for a real person and whether they stay engaged with the AI agent. Run the same call flow with different voices, and you can get different outcomes.
That can make poor performance surprisingly difficult to diagnose. Nothing obvious has gone wrong. The routing worked, the right information was captured, and the model handled the conversation correctly. Still, callers may disengage, ask for a person or abandon the call more often than expected. The usual metrics show you what happened but not necessarily why. When everything else appears to be working, the voice itself is easy to overlook.
The Decision You Made By Accident
Break a voice down, and you find a series of choices. Accent signals geography and familiarity. Pace can create calm or urgency. Formality helps determine whether you sound like a law firm or a lifestyle brand. Warmth changes whether an interaction feels personal or transactional. Gender also carries assumptions, whether we want it to or not. Together, these choices tell the caller something about your business.
Now compare that with your brand. A luxury brand answering with a generic synthetic voice. A proudly regional business speaking in an accent with no connection to its customers. A company built around personal service delivering cold, clipped efficiency.
Customers may never consciously identify the mismatch, but they feel something is off. That’s a poor first impression to create at scale.
The Uncomfortable Part
Voice design also runs directly into bias; pretending otherwise doesn’t make it disappear. The better approach is to acknowledge that these perceptions exist and make decisions based on your customers, your context and the experience you’re trying to create.
People respond differently to accents. They make assumptions and carry expectations. A voice that sounds reassuring to one person may sound distant to another. There’s no neutral default that avoids this. What matters is making the choices consciously and testing how they land with the customers you actually serve.
Testing It Like You Test Everything Else
Voice selection shouldn’t be treated as part of product and brand design. Marketing teams routinely A/B test subject lines, landing pages, calls to action and creative. There’s no reason voice shouldn’t be treated with the same discipline.
Does a different pace affect how often callers complete an interaction? Do people interrupt one voice more frequently than another? Does a warmer tone work better for one type of inquiry but slow down a straightforward transactional call? At what point does conversational become too casual? These are testable questions, and the answer shouldn’t be, “We liked this one best in the meeting.”
This is one of the areas we’re exploring through our research: looking at how characteristics such as tone, pace and personality shape conversations and how businesses can make more intentional choices about the voices representing them. The interesting part is finding the right voice for the brand, customer and conversation.
Where To Start
Start by auditing your current AI voice against your brand. Listen to it as a first-time customer would. Does it actually sound like the company you claim to be?
If your brand guidelines say warm and approachable, but your AI sounds like a satnav, you have a disconnect, and better routing or a more capable model won’t fix that. The problem isn’t what the system knows or can do. It’s how the experience feels.
Go back to the original question. If your AI’s voice were a person, would you put them on your front desk, answering customers on behalf of your business every day? If the answer is no, you probably shouldn’t let the AI do it either.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?