Voice, Chat, Email: Which Support Channel Should You Focus on in 2025?

Most companies use several support channels and AI solutions to communicate with customers. You probably do too, but unless you’re focused on using each of them to serve customers in the way they expect, you might be running into some common issues.

Take a telehealth and pharmacy provider serving patients of all ages, for example. Three different customers with three different problems might call in using three different channels:

1. A customer calls voice support because a charge feels wrong, and they want to talk it through. They don't trust chat or email for something this serious, but they hit a phone tree, can't get to a human, and hang up frustrated.
2. Another customer uses a chat window during their lunch break when they realize their prescription hasn't shipped. They expect a quick answer, but they get distracted, and when they come back, the session timed out. Now, they have to start over.
3. A third customer emails about a medical device that stopped working. They attach a photo and provide a detailed description of the issue. They're not in a hurry, but they expect a complete, thoughtful reply.

Each of these customers chose their channel for a reason. Voice was chosen when a customer wanted their emotions to be heard about urgent issues. Chat was selected for speed and convenience, while email was chosen for detail and completeness.

But if you've built all three channels with the same logic—the same AI automation, same escalation rules, same intent handling—then at least two of them will fail in each scenario because your design doesn't match what the customer came there to do.

We conducted extensive research to determine what is happening across voice, chat, and email. What we found confirms you should have a multichannel strategy, but more importantly, we found specific ways to optimize each channel so you're giving customers what they want.

What do customers want from each channel?

We surveyed both sides of the support coin (CX leaders and the customers they serve) about their actual behaviors. Our 2025 AI in CX Benchmark Report surveyed 642 U.S.-based mid-market companies to examine how businesses use AI across customer support channels and the impact it has on deflection, CSAT, and cost. The report breaks down performance benchmarks by company size, ticket volume, industry, helpdesk platform, and AI maturity, with a focus on how different automation types (like agentic AI vs. RAG or decision trees) perform in real-world use.

To complement that data, we conducted two consumer-facing "Voice of the Customer" surveys, each fielded to over 1,000 U.S. adults in early 2025. The goal was to understand how customers experience support across channels, including what they use most, which they trust, and how their expectations shift by use case, emotion, or generation. These surveys revealed preferences, frustrations, and expectations to contextualize business adoption trends with real user behavior.

The business side shows that nearly every company offers email (92%), three-quarters offer voice (75%), and over two-thirds offer chat (69%). The most common pairing is chat and email (80%), followed closely by email and voice (75%).

But while adoption is broad, usage isn't uniform. Each channel plays a distinct role, depending on the context, complexity, and customer. Consumer data reveals that people don't randomly pick a channel—they choose based on what they need to accomplish and how they feel about the problem.

When companies deploy AI, they're starting to recognize this reality. Chat leads as the most critical channel for AI deployment, at 46%, followed by email at 29% and voice at 20%. However, this prioritization reflects ease of implementation rather than customer preference.

‍

The fundamental insights come from understanding why customers choose each channel and what they expect when they get there. Customer preferences shift dramatically based on use case, emotion, and even generation, and your AI strategy must account for these variations.

1. Voice is where customers want their emotions understood

Voice is still the most emotionally important support channel. When something feels personal or urgent, customers want to talk to someone.

Sixty percent of customers say that voice is the most effective way to resolve a problem, and 66% say it’s the channel where they’re most likely to express emotions such as crying, frustration, or confusion.

That emotional weight spans generations. Thirty-five percent of Boomers choose voice first, and even Gen Z prefers it in high-stakes situations, such as billing disputes and account problems. You also see this across industries where the stakes feel personal, like banking, healthcare, insurance, and food service.

Despite that, ticket volume over voice is flat. Only 31% of companies report any growth in voice support, while 69% report it staying flat or trending downward.

Our findings suggest this isn’t because customers don’t want to use voice but because companies have made it too hard to use. Rigid phone menus and lengthy wait times deter people. Forty-six percent of customers report becoming frustrated or hanging up while navigating phone systems. Thirty-five percent stay on the line but feel annoyed. Just 11% say they’re fine with the process.

But while voice carries the most friction, it also holds the most potential to be transformed with AI. An effective AI voice agent can pick up the phone immediately, doesn’t need a phone tree or a script to operate, and can be empathetic in the face of human emotion.

2. Chat is where customers want issues resolved quickly or passed to a human

Chat is rising fast for two reasons: it’s fast for customers, and it’s structured for AI. That’s why 69% of companies using AI offer chat and why nearly half say it’s their most critical channel.

It’s the easiest place to launch AI, but it’s also the easiest place to get it wrong. Customers come to chat because they expect speed, but they also need flexibility. They want to multitask, step away, and come back to a system that remembers what they said.

Nearly a third of customers report multitasking while chatting, and 46% cite the written record as one of the top benefits; however, this flexibility also creates risk. When the system loses context, the conversation resets. 51% of all customers report forgetting they were chatting, having the session time out, and having to start over. Among younger generations, the problem is even worse: 44% of Millennials and 50% of Gen Z report having to restart chats after stepping away.

If their issue isn’t resolved, a third of customers expect to be able to escalate to a human mid-chat. Another 29% expect a complete resolution in the chat window without being routed to a different channel. When the system can’t meet those expectations because the session reset or the handoff broke, it damages trust.

The AI-powered chats that frustrate customers are either not trained to take action and solve an issue themselves or don’t handle handoffs well. They’re optimized for speed, not for resilience, and the result is fast answers that don’t help or long delays when trying to route to a human agent.

3. Email is the default for many companies, but it demands smarter interactions

Email is the most common support channel offered by CX teams. Ninety-two percent of companies use AI to power their operations, but it’s often misunderstood. Most teams still treat email like a slow version of chat, even though it plays a different role in the support experience.

Customers don’t expect it to be fast, but they do expect it to be thoughtful. And they use it when they want to explain something once, clearly, without back-and-forth. They use it for account updates (11%), for checking order status (20%), and for resolving bad experiences. It’s the third most common initial contact method overall, behind only chat and voice.

It’s where customers go when their issue is too detailed, too long, or too complex for a live conversation, which is why it ranks low in terms of preference for urgency. Just 11% say it’s the most effective way to get a resolution, and only 9% use it for dispute resolution.

At the same time, email volume is growing. Forty-three percent of companies report an increase in ticket volume over email, and the largest share of email-based support teams handle between 5,000 and 10,000 tickets per month. Customers still use it because it’s sometimes the only option, or it provides them with a space to explain the full issue, send attachments, and document everything in one place.

The length and depth of email interactions mean that the tone, format, and language used in emails are unpredictable. One message might be a few words, and the next might be four paragraphs covering multiple questions.

Unlike chat, which benefits from structured historical ticket data, email requires AI that can handle long-form language, pick up emotional tone, and stay organized through multiple parts of a conversation. That only works when models are trained on real email threads—not just FAQ snippets or knowledge-base summaries.

Tips for improving each channel

AI is the solution to optimize each of these channels. Still, success depends on finding the right partner, training it properly, and setting it up to deliver what customers want from each channel.

Many companies are already making this shift. Fifty-seven percent of companies now use AI in customer support and a quarter of those started within the last six months.

Most began with chat because it's easy to implement, familiar, and low risk. But usage is expanding rapidly. As companies get more comfortable with AI, they're moving into voice and email. Sixty percent now offer all three channels—chat, email, and voice—alongside AI automation.

The difference between companies that succeed and those that struggle comes down to implementation. You can't deploy the same AI logic across all three channels and expect it to work. Each channel requires a different approach because customers bring different expectations to each one. But it requires intentional design choices that match AI capabilities to customer expectations.

Use voice AI that sounds human and takes action immediately

We've already covered that people use voice when they're feeling emotional or like something is urgent. Yet when they pick up the phone, they get an experience that seems tailor-made to frustrate them further. They're faced with interactive voice response (IVR) trees that force them to push buttons and wait before they can even speak to a human.

It’s so frustrating that people hang up after just one to two IVR button presses (11%) or three to four presses (35%). Nearly half of all customers abandon their calls before reaching a human agent, not because their problem isn't real, but because the process is broken.

AI can immediately solve this first problem—it can easily pick up the phone immediately, eliminating phone trees entirely. But answering fast isn't enough. The AI must convincingly demonstrate empathy in the face of an emotional human, as people choose voice specifically when they want to be understood.

It must also be able to fully resolve a customer’s issue. When we asked customers which channel is most effective for getting issues resolved, voice dominated at 60%, followed by chat at 19% and email at 11%. That last requirement is where most voice AI systems fail because they demand agentic capabilities.

The foundation of successful voice AI is system integration. You need extensive API connections to every system your human agents use. Customer databases, billing platforms, inventory management, shipping systems, CRM tools—if your AI can't access and modify these systems in real-time, it becomes just another frustrating dead end for emotional customers.

Forethought Voice is built specifically to meet these requirements. It eliminates phone trees entirely, uses advanced voice synthesis that sounds genuinely human, and leverages our Autoflows technology to take real action through your existing systems. We've invested heavily in making it sound natural and empathetic, but the real breakthrough is that it can access your tools and resolve issues immediately.

[Embed Video]

Most voice AI solutions fail because they only solve part of the problem—they might sound conversational, but they can't actually resolve issues, or they can integrate with some systems but sound robotic when customers are emotional.

Forethought Voice is designed to deliver all three requirements together: fast phone pickup, genuine human-like conversation, and deep integration with your existing systems to take immediate action. The companies succeeding with voice AI aren't settling for partial solutions that force customers to still transfer to humans for actual resolution.

Optimize chat to solve issues instantly and hand off to humans seamlessly

Chat is where most companies start AI deployments because it's the easiest place to implement and test. Plus, they know that customers expect chat to be lightning-fast at solving their issues, and AI excels at speed. But customers don't give chatbots many chances to prove themselves.

When a chat AI can't immediately solve an issue, customers expect to be transferred to a human agent. That means that if your AI-powered chatbot cannot actually resolve issues, most conversations will still escalate to humans anyway, defeating the purpose of automation.

Again, the most effective chatbots are agentic. They need to take action and solve customer problems by integrating with your current systems so they can perform the same tasks your human agents can—processing returns, updating accounts, checking order status, and applying discounts. Without these capabilities, you're just creating an expensive screening tool that frustrates customers.

But chat has a unique challenge that voice doesn't: customers treat it as asynchronous communication. The data shows that 46% of respondents value chat for the written record, and 31% want to multitask while chatting. They're treating it more like email that happens to be faster.

This means that, for chat to work the way customers expect, AI-powered chatbots must remember and pick up on old conversation threads with returning customers who got distracted and abandoned previous chats. When someone comes back three hours later asking, "What was that link you sent me?" your AI needs to understand the full context and continue the conversation.

Forethought Solve is designed specifically for these chat requirements. It's an omnichannel AI agent that can resolve customer inquiries from initial contact through final resolution. Unlike basic chatbots that simply answer questions, Solve utilizes our Autoflows technology to take real actions, such as processing refunds, updating shipping addresses, applying for account credits, and handling complex, multi-step workflows.

[Video Embed]

The system maintains full conversation memory across sessions and channels, so customers never have to repeat themselves or start over. If someone begins a conversation in chat, abandons it, then returns later or even switches to email, Solve picks up exactly where they left off with complete context.

Solve also handles the seamless handoffs that customers expect. When escalation is needed, it transfers customers to human agents with full conversation history and context, eliminating the frustrating "let me review your case" delays. The AI continues learning from these interactions to handle similar issues autonomously in the future.

Treat email like a specialty channel for complex issues

So many companies use email to power support because it's the easiest channel to set up and automate. But despite being the most popular channel offered by businesses, it's definitely not the most preferred by customers. Only 9-20% prefer email across different scenarios, despite 92% of companies offering it.

But when customers do choose email, it's because they're dealing with complex problems. They want a place to attach screenshots, receipts, and other documentation. They need to lay out all the information in one place and expect a comprehensive response that addresses every aspect of their issue.

Email conversations are fundamentally different from other channels because they’re lengthy, detailed, and multi-layered. Customers frequently shift tone within a single message, present multiple problems simultaneously, and include extensive background information that AI must parse and understand completely to be successful.

If you're going to power your email interactions with AI, the most important thing you can do is train it on actual email interactions from your support team—not chat transcripts or external data sources. Email communication has its own patterns, language, and customer expectations that are distinct from real-time channels.

Forethought's email solution is specifically designed for this complexity. Rather than treating email like slow chat, our AI is trained on full email case histories to understand the nuanced, detailed communication style that email customers expect. The system can parse lengthy messages with multiple issues, maintain context across long threads, and provide comprehensive responses that address every concern raised.

The technology also handles the documentation aspect that makes email valuable. It can reference attachments, integrate with your systems to pull detailed account information, and create complete resolution records that customers can save and reference later.

Treating email like a specialist channel. Most companies over-invest in email automation because it's easy to measure response times, not because customers want rapid-fire email exchanges. When someone chooses email over chat or voice, they're signaling they want a complete, thoughtful response that fully resolves their issue, not a quick deflection to another channel.

The future of customer support isn't choosing between channels

When AI handles the routine work correctly in each channel, your human agents can focus on what they do best: solving complex problems, building relationships, and creating the moments that turn frustrated customers into loyal advocates.

The companies getting this right aren't just using AI to deflect tickets or reduce cost; they're also leveraging it to enhance customer experiences. They're creating support experiences that customers prefer over traditional methods.

AI can give customers exactly what they want from each channel if the right technology is implemented thoughtfully. Ready to transform your customer support from a cost center into a competitive advantage? See how Forethought can optimize each of your support channels for what customers actually want.

‍