Finding your (AI) voice…..

Would you trust AI to make a phone call on behalf of your business?

 

For some business owners, that question alone will cause fear, but for others it will spark their curiosity.

 

We all know by now that AI is out of the bottle and is not going back in. But AI is more than just ChatGPT and Claude - and one of the newest emerging elements of the technology is AI voice agents that can make external calls on behalf of your business.

 

You can train them to say what you want them to say, you can choose how they sound, and you can use them as you please… but will it work for everyone?

 

Using AI voice

I’m often asked which businesses AI voice works for and which it doesn’t. Essentially, AI voice can work for pretty much any business. But I think the better question is about what kind of conversation you’re trying to get the AI to have.

 

If it’s a fairly straightforward conversation where a prospect wants information, needs to be qualified, or wants to book something, then AI voice can be a very effective solution.

 

For example, take businesses like air conditioning, roofing, insurance, med spas, gyms, dental, automotive, restaurants, or most other service businesses. A potential customer has made an enquiry; they want some questions answered, and they potentially want to book an appointment. There is not necessarily a huge emotional relationship needed at that point.

 

A restaurant is also a good example. If I phone a restaurant, I probably just want to know if they have a table available, what time I can book, or whether they can accommodate something specific. If an AI can answer those questions and actually put the reservation into the booking system properly, I am completely happy speaking to it.

 

The problem is when the AI is clunky and asks you for information but then can’t understand what you are actually telling it. Everyone hates that … but not necessarily because it’s AI. They hate it because it’s slow, frustrating, and completely useless.

 

That is the big distinction. Most people who say they don’t like AI voice have usually just experienced badly implemented AI voice.

 

There are a few things needed for AI voice to work properly. Firstly, the voice actually needs to sound good, have a nice tone and be pleasant to converse with. Then, the latency needs to be really low, so there is no big pause between the person speaking and the AI responding. If it takes a second or two every time, people start talking over it and the whole conversation just feels broken.

 

The voice itself also needs to have some personality. It needs inflection, changes in tone, and use of natural filler words where appropriate. Even small things like saying “um,” “okay,” or “yeah, that makes sense” can make the conversation feel much more natural for the customer.

 

The speaking speed matters as well - a lot of AI voices speak far too slowly, and it sounds like they are reading from a script one word at a time. We normally want it speaking slightly faster, because that’s closer to how people actually speak on the phone.

 

It also needs to handle interruptions properly. Real people don’t wait patiently for each other to finish every sentence. They interrupt, change direction, correct themselves, and ask questions halfway through an answer. If the AI falls apart every time that happens, it’s not going to work.

 

Emotional connection

Where I’d suggest companies need to be careful with AI voice is when there’s a strong emotional connection needed between the customer and the business or the person they are speaking to.

 

For example, if somebody is looking for therapy, or perhaps even a funeral director, as the technology currently stands, I’d recommend against using an AI to call them and try to manage that conversation. That person needs to feel understood and cared for during an emotional time, and having an AI handle the relationship could feel completely wrong.

 

I would also warn against using AI voice as a replacement for ongoing customer relationships. If a client is paying for a high-value service and needs to speak to their account manager or project manager, there is a level of trust involved that has been built up over time. They want to know there’s a real person who understands their business and is responsible for helping them.

 

Lead management

Where AI voice does come into its own is lead management.

 

For example, when somebody has just filled in an online form, there’s no relationship established at that point. They normally want a fast response, some useful information, and a clear next step. That is where AI voice can work incredibly well. It can contact the lead immediately, qualify them, answer questions, and book them an appointment without somebody on the team needing to be available.

 

There are some key considerations when working with AI voice agents.

 

Firstly, how knowledgeable the AI is. It needs a comprehensive knowledge base above all else, including clear instructions about what it can and cannot say, and rules for when it should transfer the conversation to a human. This is to avoid it getting things wrong, giving out incorrect information or not representing your business in the way you want it to. 

 

Then there is the integration side, which is probably one of the hardest parts to get right.

 

Going back to the restaurant example, there is no point having an AI that sounds amazing if it can’t see the actual table availability or write the reservation into the restaurant’s booking system. It might have a brilliant conversation, but if the booking does not exist when the customer arrives, the whole thing has failed. So the AI technology has to be fully integrated into the system it needs to use.

 

Similarly, when following up with leads, the AI needs to know which calendar it can book into, update the CRM properly, add the right notes and tags, route the lead to the correct person, and record exactly what happened on the call.

 

That’s why I think one of the biggest mistakes business owners make is focusing far too much on the voice itself.

 

Yes, the voice matters, but it’s only the part the customer hears. The thing that makes the system actually useful is everything happening behind it: the knowledge, prompting, integrations, guardrails, routing, escalation rules, and ongoing quality control.

 

Finally, many AI voice agents these days sound like humans - in fact, in tests, often people don’t even realise they’re speaking to an AI and not a real person. However, ethically, companies must make customers aware that they are speaking to an AI voice agent and not a human. Some businesses might think they’re being clever by not letting on that they’re using AI in this way, but the right thing to do is to make sure that customers are fully informed about this - that alone will help build trust with them.

 

So, in simple terms, in my view, AI voice works really well when the conversation has a clear objective and doesn’t require a deep emotional relationship. It is especially strong for lead follow-up, qualification, booking, FAQs, and basic customer requests.

 

It works less well when somebody needs genuine emotional support, complex human judgement, or an established relationship with a person they trust.

 

And in most cases, the question is less about whether AI voice can work for the business and more about whether it has been implemented properly. 

 

---

 

Arin Darcan is the founder of Team Follow Up AI, specialising in the implementation of AI voice agents for businesses

 

Main image courtesy of iStockPhoto.com and Ole_CNX