New Consumer Research: How Consumers Grade AI Support on Their Most Complex Issues

Key takeaways

• 90% of U.S. consumers had at least one complex support issue in the past year
  
  • 50% of them had at least one complex issue that was never resolved
• 6 in 10 trust AI for simple requests, but not complex ones
• 68% would abandon AI support after two failures or fewer
• 50% got generic, unhelpful answers from AI on their most recent complex issue
• 64% trust AI more when they control when to reach a human
• 59% judge AI support on whether it resolved the issue completely

Consumers will give AI a chance on their hardest support problems, and they will take that chance back quickly. In a survey of 1,040 U.S. consumers, conducted by CMSWire and commissioned by Zingtree, 68% said they would abandon AI support after two failures or fewer.

Most support teams have already automated the simple, high-volume requests: password resets, balance checks, order tracking. Complex support is the next push, so we asked the people on the receiving end how it is going. We wanted their view because internal numbers like containment and deflection can look healthy while customers are still stuck.

In this survey, a complex support issue is one that meets at least one of three conditions:

• It required multiple interactions
• It took more than 30 minutes to resolve
• It involved a high-stakes matter, such as a financial or healthcare issue

This post shares nearly every number in the research. The full report, Beyond Basic CX: How Consumers Grade AI Support on Their Most Complex Issues, has the charts and segment breakdowns. Download the full report.

How common are complex support issues?

Nearly every consumer has one. 90% of U.S. consumers had at least one complex support issue in the past 12 months, and nearly 40% say more than a quarter of all the support issues they raised last year were complex.

Many of those issues end without a fix. Among consumers with a complex issue:

• 50% had at least one issue that was never resolved
• 1 in 4 abandoned a support issue in the past year
• 33% spent multiple days trying to resolve a complex issue before giving up

Which support channels fail on complex issues?

All of them. Even the ones staffed by people. When consumers described a complex issue that was never resolved, these are the channels they had tried (respondents could pick more than one):

• Live chat with a human agent: 50%
• Automated chatbot: 44%
• Email to customer support: 44%
• Phone support with a human agent: 37%
• Phone support with an automated system: 34%

Live chat with a person tops the list. Complex issues, and the failures that come with them, show up in every channel.

Why do complex support issues go unresolved?

Consumers mostly blame the process. Asked why their complex issue was never resolved, they said:

• 37%: the company could have solved the problem but made the process too difficult or slow
• 34%: the company understood the problem but didn't know how to solve it
• 19%: the company never understood the problem

The AI-specific failures follow the same pattern. Among consumers with a complex issue:

• 50% got generic, unhelpful answers from AI during their most recent complex issue
• 50% had to repeat information to a human after being transferred from AI

The report traces these to three failure points: AI that doesn't understand the issue, no transfer to a human when AI can't resolve it, and transfers that lose the context of the issue.

How many chances do consumers give AI?

Two. 68% of consumers would abandon AI after two failures or fewer.

The biggest single frustration in complex support is wasting time with AI that fails to resolve the issue (27%). Second is repeating information to a human after AI fails (24%). And consumers who have had an unresolved complex issue are more likely to distrust AI with the next one.

That distrust is already visible:

• 6 in 10 consumers trust AI for simple requests, but not for complex issues
• 44% actively try to bypass AI support whenever possible
• 67% believe companies design customer support around their own cost savings and efficiency

Does trust in AI on complex issues vary by industry?

Yes, and insurance faces the most skepticism. Here is the share of consumers who agree that "AI is useful for simple issues but I don't trust it for complex ones," by the industry where their complex issue went unresolved:

• Insurance: 68%
• Telecom: 65%
• Travel: 64%
• Technology: 63%
• Retail: 61%

What makes consumers willing to use AI on complex issues?

Control over reaching a person. Consumers will start with AI when they know they can get out:

• 64% trust AI more when it gives them control over when they can transfer directly to a human agent
• 64% say a good AI-to-human transfer makes them more willing to use AI again
• 79% of daily AI users are willing to use AI for complex support if they can easily reach a human when needed
• 45% prefer to start a complex issue with AI if it offers easy access to human support
• 60% don't mind interacting with AI when it resolves their issue

Even consumers who would rather talk to a person accept AI behind the scenes. Of the 66% who prefer human agents to AI, those consumers say they are comfortable with agents using AI tools to assist them.

Does personal AI use change how consumers treat AI support?

Yes. Consumers who use generative AI in daily life are more receptive to AI support, and the rare users route around it. 61% of consumers who have used a generative AI tool fewer than three times try to bypass AI in customer support whenever possible.

When AI gets simple issues right, customers are more willing to bring it a complex one. Each AI interaction, simple or complex, sets the terms for the next.

How do consumers judge whether AI support worked?

By the outcome. Asked what matters most when judging whether AI customer support was successful:

• It resolved my issue completely: 59%
• It understood what I was asking: 59%
• It resolved my issue quickly: 50%
• It gave accurate information: 50%
• It transferred me to the right human when needed: 49%

Who or what solves the problem matters less to consumers than whether it gets solved.

What should CX leaders do with these findings?

Our work in high-stakes, highly regulated industries has shown that AI's role in resolving complex issues is fundamentally different from handling simple, straightforward queries. Consumers are willing to give AI a chance on complex support, and the window to get it right is small. If AI can't resolve an issue, or fails to transfer the customer to a human agent with full context, a business risks the trust and loyalty it has worked to build.

The research points to four moves:

1. Tell customers when AI is part of the support process. Trust starts with knowing who you are talking to.
2. Hand off to a human as soon as AI reaches its limit. A well-timed transfer is a successful outcome, and 64% say a good one makes them more willing to use AI again. We wrote about how to set that line in Human-in-the-Loop AI: Setting the Autonomy Dial for Every Customer Conversation.
3. Send the full context with the customer. Half of consumers had to repeat themselves after a transfer, and it is their second-biggest frustration.
4. Measure resolution from the customer's side. Containment and deflection rates can look healthy while customers abandon the issue. Judge AI the way your customers do: did it understand, and did it resolve?

I wrote more about what these findings mean for agentic AI in Why Complex CX Is the Real Proving Ground for Agentic AI on CMSWire.

Every finding at a glance

Methodology

The Customer Support Journey survey was commissioned by Zingtree and conducted by CMSWire INSIGHTS, the research arm of CMSWire, from July to August 2026. The online survey sampled 1,040 consumers aged 18 and over who live in the United States and had experienced a complex support issue in the past 12 months. Unless noted, percentages refer to those consumers.

• Age: Gen Z (18–29) 25%, Millennials (30–45) 35%, Gen X (46–61) 32%, Boomers and Silents (62–80+) 8%
• Gender: Female 54%, Male 46%
• AI usage (how often they use a generative AI tool for personal or work purposes): High (daily) 32%, Moderate (weekly) 26%, Low (monthly or yearly) 15%, Rare (twice or fewer) 27%
• Household income: Under $49,999 44%, $50,000–$100,000 34%, $100,000–$149,999 13%, $150,000–$250,000 7%, Over $250,000 2%
• Region: South Atlantic 25%, East North Central 16%, West South Central 14%, Middle Atlantic 12%, Pacific 10%, East South Central 9%, West North Central 5%, Mountain 5%, New England 4%

How to cite this research: Zingtree and CMSWire INSIGHTS, Beyond Basic CX: How Consumers Grade AI Support on Their Most Complex Issues, October 2026. Survey of 1,040 U.S. consumers, July–August 2026.

Frequently asked questions

What counts as complex customer support?

An issue that required multiple interactions, took more than 30 minutes to resolve, or involved a high-stakes matter such as a financial or healthcare issue. 90% of U.S. consumers had at least one in the past year. The full report breaks down how those issues played out.

Do consumers trust AI for complex customer support?

Partly. 6 in 10 consumers trust AI for simple requests but not for complex issues, yet 64% trust AI more when they control when to reach a human. Trust follows the handoff. Read how to design that handoff in Human-in-the-Loop AI: Setting the Autonomy Dial.

How many times will customers let AI fail before giving up?

Twice. 68% of consumers would abandon AI support after two failures or fewer. More on why the window is closing in Why Complex CX Is the Real Proving Ground for Agentic AI.

Does transferring a customer from AI to a human mean the AI failed?

Consumers count a good transfer as a good outcome: 64% say a smooth AI-to-human transfer makes them more willing to use AI again, and 49% judge AI support on whether it sent them to the right human. What damages trust is a transfer that loses context, which happened to 50% of consumers. See Human-in-the-Loop AI for how teams decide when AI should hand off.

What do customers want from AI customer support?

Resolution and understanding. 59% judge AI on whether it resolved the issue completely and 59% on whether it understood what they asked, ahead of speed (50%) and accuracy (50%). For how Zingtree keeps AI accurate on these issues, read How We Keep AI From Guessing in Complex CX.

Who conducted this research?

CMSWire INSIGHTS, the research arm of CMSWire, conducted the survey for Zingtree from July to August 2026, with 1,040 U.S. consumers. Download the full report.

Get the full report

Beyond Basic CX has every chart from this research, with the breakdowns by AI usage and industry. Use it to make the case for better handoffs and full resolution before AI takes on the issues your customers care about most.

Download Beyond Basic CX