Good Customer Support Wins Back Customers, But is AI Delivering on That?

Customer support has long been a place where a company can win back someone it has let down, but recent developments with AI may be weakening that role, new RepTrak data shows.

In our Q3 2026 research, 60% of the Informed General Public said a company that handles problems or complaints well can improve their opinion of it, even after something went wrong. Only 35% said AI-powered customer support is welcome, and that was with the benefit written into the question: when it makes getting help faster, clearer or easier.

Even on those terms, about two-thirds held back: 44% kept a neutral opinion, 15% disagreed with the statement and 6% weren't sure. A likely reason shows up in the same question: Most people (59%) say a company damages its reputation when customer support feels designed to stop them reaching a real person.

People will accept AI that helps them, but they hold it against a company when it keeps them from a person.

Business surveys, meanwhile, point the other way. More than 75% of customer service leaders say executives are pushing them to adopt generative AI, according to Gartner's December 2024 survey of service leaders. In Salesforce's 2025 State of Service research, service leaders expected AI agents to cut service costs and resolution times by about 20%, and AI climbed from 10th to 2nd on their priority list. The push is coming from the top, and it's framed largely around cost and speed.

That puts the moment most likely to repair a relationship in the hands of the approach people are least willing to accept.

People agree on what good support does. AI is the exception.

Set against the rest of the question, the AI result stands out. The four other statements drew between 59% and 65% agreement, and none drew more than 4% disagreement. Across all five, AI support was the only one without majority agreement, and the only one with real opposition. These are separate answers to the same question, and we haven't yet cut one against the other, so we can't say how many of the 60% sit in the 35%. The size of the gap is clear enough without that.

Customer service is the AI benefit people believe in least

The Q3 gap fits a pattern in our tracking. Since Q3 2023, our Current Events research has asked about seven statements on AI. In Q2 2026, "AI improves customer service" drew the least agreement of the seven globally (29%) and the most disagreement (18%). More people agreed that AI has a negative impact on society (31%) than agreed it improves customer service. The other two benefits we track, making life easier and helping businesses run more effectively, each drew 33%.

The picture is starker in the markets where most of our readers work. Of the 35 markets in that study, customer service was the lowest-rated AI statement in 14, including the US, the UK, Canada, Australia, Germany and France. In six, more people disagreed than agreed. In the US, 19% agreed AI improves customer service and 29% disagreed, and Americans were nearly twice as likely to agree that AI harms society (34%). Markets such as India, China, Saudi Arabia and Brazil, where agreement runs from about 49% to 58%, lift the global figure.

People still want companies to get service right. In Q1 2026, improving the customer experience was the second-highest priority people set for companies (36%), just behind data privacy and cybersecurity (37%). AI came last of nine, at 16%. People are asking for the outcome and ranking the tool at the bottom.

The people who shape opinion see it differently. In our Q4 2025 wave, which covered 11 markets, 43% of Influencers agreed AI improves customer service, against 29% of the Informed General Public. A team that takes its read on AI support from industry conversation risks hearing a warmer story than its customers would tell.

What wins people back is specific

The Q3 statements that drew near-consensus describe what recovery requires, and they read like a design brief. Nearly two-thirds (65%) say good support means resolving the issue quickly without making them repeat themselves. Three in five (62%) say companies should make it easy to get help through the channel that suits them best.

Neither statement mentions AI, and both describe points where AI support can help or fail. "Without making me repeat myself" names the handoff from an automated system to a person, when a customer can be forced to start over. "The channel that suits me best" makes AI one route among several. A system that carries the customer's context through the handoff, and sits alongside other channels, meets both.

That's why the 59% matters. Put it beside the 60%, and support built to keep people from a person does two kinds of harm: it gives up the chance to win a customer back, and it adds a grievance of its own.

Our Q1 2026 research shows the same line from another side. People use AI most as help, at work and for research, where 33% to 37% use it at least weekly. They use it least to make decisions: 25% use it weekly to compare products before buying, and 28% never do. People are comfortable with AI that assists them and much less comfortable handing it the decision. A support system that decides whether a customer gets to a person has made that decision for them.

In our model, customer support is a direct experience touchpoint, alongside the buying experience. In this year's Global RepTrak 100, we noted that Products & Services and Performance are the drivers most legible through direct experience. Products & Services was also the only driver to fall in 2026, down 0.1 to 76, its first decline after consecutive years of gains, led by a drop in the high quality products and services factor. We can't connect that decline to support, let alone to AI support. It does mean the driver people read most directly from experience is the one that slipped this year.

What to do before AI support scales

Support is where people say a company can earn its way back, and AI is the part of support they're least ready to accept. The two can coexist if AI handles routine requests and a person handles the moments that decide whether a customer comes back.

That makes an AI support rollout a reputation decision. In practice, that means four steps:

1. Route recovery moments to people.

   Let AI take routine requests, and send complaints, repeat contacts and anything that follows a failure to a person quickly. Those are the moments the 60% describes.

2. Before launch, walk the journey against the standard people named.

   Can a customer resolve the issue without repeating themselves, including when the system hands off to a person? Can they get help through the channel they chose? Is a person reachable from every step, and does the customer know how?

3. Read the results by market.

   In the US, the UK and Canada, more people doubt that AI improves customer service than believe it, so a rollout that works in one market can cost you in another.

4. In support-page copy and customer communications about the change, lead with the outcome:

   faster answers, and a person when you need one. Our Q4 2025 guidance made the same recommendation: concrete, human-centered outcomes, paired with visible commitments on oversight.

RepTrak measures how stakeholders perceive a company across seven drivers and 23 factors, with touchpoint reach and impact showing which channels, customer support among them, are helping and which are hurting. That's how a team can see whether a change to support moved perception, and in which market.

If you want to know whether your support experience is winning customers back or pushing them away, talk to us.