You’ve just assembled your new grill, but you’re stumped on how to hook it up to the gas line. So you call customer support and start talking with an agent. Turns out they’re an AI, not a person, but it’s hard to tell the difference.

Answering questions in conversational English, the agent asks you to share a video of the barbecue so they can “see” it — then talks you through a fix in real time. Problem solved.

Sound like science fiction? Soon, such calls will be routine.

That’s good news for consumers. When it comes to AI and customer service, we’re going to see a virtuous cycle in 2024. As the technology gets better, cheaper and easier to use — a far cry from the stiff, robotic chatbots of just a year or two ago — more companies will embrace it.

In turn, many people will have their first truly positive experience with an AI agent. Consumers are ready to engage. In a recent poll, eight out of 10 respondents said they’re comfortable with using generative AI for customer support or service inquiries.

As the founder of a company that specializes in AI customer service, I know how much companies value this technology. Already, AI agents handle upward of 70% of customer conversations, according to one estimate. As those agents start outperforming their human counterparts in many respects, look for the quality of customer service to greatly improve.

With that in mind, here are four trends that will drive a better AI customer experience in 2024.

Related: Why AI Is the Next Revolution In Customer Service

1. Generative AI for voice support goes mainstream

“For billing, press 8.” We’ve all experienced the frustration of getting stuck in a phone menu with an automated voice agent whose only skill seems to be pressing our buttons. Good news for consumers in 2024: Those agonizing calls will soon be history.

That’s because customer support is shifting from declarative to generative AI. For users, it’s the stark difference between punching numbers and having a real conversation about why that mystery charge appeared on your credit card statement. Even better, rather than direct you to the billing department, the agent will give you an informed answer about what happened and even be able to resolve the issue on the spot.

It’s no leap to say that some of these experiences may soon prove more effective and satisfying than traditional interactions with human agents. Often, through no fault of their own, the people who staff support lines could use better training, don’t speak the same first language as their customers and lack access to the business systems of the company they represent.

The new breed of AI agents don’t face such challenges. Capable of absorbing and following detailed instructions, they’re fluent in your language of choice. These agents also play nicely with company systems, allowing them to handle tasks like tracking orders and issuing refunds. The most capable ones will have the same authority as a manager, removing the need to pass a customer between people to resolve an issue.

For companies, the benefits are real. In one study, businesses using generative AI in contact centers tripled customer satisfaction versus peers that didn’t.

2. Multimodal communication takes root in customer support

In the new and improved world of AI customer support, voice is just one element of truly multimodal communication, which will spring to life this year.

Traditionally, phone support, chat support and other channels like email have been largely siloed. We’ve probably all had the frustration of just wanting to text a photo or video on a service call, only to find there’s no easy way to do it. Even as AI has transformed customer service, these platforms have remained largely walled off from one another. Until now.

In 2024, customers will be able to contact a brand via their channel of choice, moving seamlessly between talk, text, video and pictures, just like they do in chats with friends and family. All of those channels will work together to help them best resolve an issue.

Multimodal will give AI customer service agents the power to do everything from eyeballing a broken product to offering customized financial advice. Can’t figure out how to mount that new OLED TV? Start a video chat with an AI agent, which can use machine vision to pinpoint exactly the mounting bracket you need, then text you a code to buy one.

We caught glimpses of this fast-approaching multimodal future in Google’s recent demo of its new Gemini AI model. The model is ostensibly a breakthrough, powering a chatbot that holds spoken conversations with users and can recognize, distinguish between and describe images. Yes, Google took heat for making it look like Gemini could process visual information instantly, but it won’t be long before multimodal customer support will be table stakes for any business using AI agents.

Related: Will Artificial Intelligence Replace Human Interaction? 4 Ways AI is Impacting and Empowering Customer Experience

3. Brands peer inside their AI agent’s brain

For all of the upsides, some early efforts at generative AI for customer support have left companies and consumers frustrated. Brands have struggled with irrelevant or unhelpful answers, “AI hallucinations” and even bots that defy business logic. Just ask the car dealer whose AI agent offered one customer a brand-new Chevy for $1. The classic refrain: “Boy, are these bots dumb.”

What’s often overlooked is that poor outcomes are usually the result of poor or unclear instructions from humans. Until now, however, it’s been exceedingly hard for companies to peer inside their AI’s decision-making process and understand the “why” behind decisions — the frustrating “black box” problem. But new software tools are shedding light on the reasoning that informs AI responses, giving brands a chance to see where things went wrong and to correct course.

That makes a big difference in customer service, where understanding why an AI agent made a decision is just as important as the decision itself. In fact, it’s the only way to drive improvement.

Let’s say an agent gave a customer a refund that wasn’t warranted. By using a tool to examine the agent’s reasoning, the company discovers that the person who trained it gave conflicting instructions. What looked like a mistake on the agent’s part was actually a management error. Knowing that, the company can issue new guidance to avoid future mix-ups.

Like people, AI agents often have a good reason for a questionable decision, given the chance to explain themselves.

4. AI manager becomes a hot job

The growing capacity to course-correct AI agents hints at a broader trend in 2024. Smart companies are increasingly treating their AI agents not like a piece of software but like an actual employee. Brands seeing the greatest return from their AI investments have quickly discovered that it’s not enough to simply buy an agent and set it loose. To thrive, that worker needs to be onboarded, measured and coached — like any other employee.

And those tasks call for a new breed of manager we’ll see more of in 2024.

AI employee manager is already a key role at companies like money management firm Wealthsimple, one of our clients. Along with setting goals for their AI agents, giving them feedback and helping them learn, AI managers must bring data analysis, project management and technical skills to the table. It represents a true leadership role — one with the potential to make an outsized impact on business results — and will become an important career stepping stone for young professionals across industries in 2024.

Related: 5 Strategies to Protect Your Company’s Customer Service from AI Pitfalls

Sooner rather than later, AI agents will start collaborating with each other, too. For example, a customer service AI agent inside a streaming platform might notice that a login button is broken for iOS users. Time to alert the code generation AI agent to issue a fix. These generalist and specialist agents will need a new class of “HR” to coordinate them. At first, that job will likely fall to the CIO, who will turn to new dashboards to track the AI team’s performance, analogous to performance management software for people.

The biggest transformation of all in AI customer service this year may well be our own expectations. Early AI support experiences may have underwhelmed some of us. In 2024, prepare to be wowed.