I've been thinking about this question for months. We build support tools, so we talk to a lot of support teams. And something has shifted. The conversations used to be about "how do we handle more tickets?" Now they're about "why are our customers so angry?"

The frustration is real. It shows up in the data, in the conversations, and in the Reddit threads where people vent about trying to cancel a subscription or get a refund from a company whose support page leads to a chatbot that leads to a dead end.

So let's look at what's actually happening.

The data says yes

Forrester has tracked customer experience quality across hundreds of US brands since 2016. Their CX Index measures three things: effectiveness (did you get what you needed?), ease (was it hard?), and emotion (how did it make you feel?).

All three dimensions have declined. CX quality fell to a new all-time low of 68.3 in 2025, its worst score since the index began in 2016. In 2024, 39% of brands had statistically significant CX declines, the worst year in the index's history. In 2025, for the second year in a row, 25% of brands declined while only 7% improved. The rest were flat. The direction is unmistakable: it's getting worse, not better.

The American Customer Satisfaction Index tells the same story from a different angle. Customer satisfaction dropped to 77.3 out of 100 in Q4 2024, then fell further to 76.9 in Q4 2025, where it has stagnated. On an annual basis, the ACSI has fallen 0.5% and "has not materially increased since 2017." More than a decade of supposed innovation and the needle hasn't moved.

These aren't obscure metrics from biased sources. Forrester and ACSI are the gold standards of customer experience measurement. When both say the same thing, it's real.

What's causing it

There's no single cause. It's a combination of forces that reinforce each other.

Companies are spending less on support relative to revenue. As companies scale, they pressure support teams to handle more with less. The mantra is "do more with less" and the metric is cost-per-ticket. When the primary goal is cost reduction, quality suffers. This isn't a 2024 problem; it's been building for a decade. But AI gave companies a new tool to accelerate the cost-cutting, and many used it aggressively.

Phone support is disappearing. Major companies have eliminated or buried their phone numbers. Airbnb, Uber, many SaaS companies, and increasingly, banks and airlines. The Washington Post ran a piece titled "Vanishing phone customer support is driving us all insane," which resonated because everyone has experienced the frustration of not being able to call a company that used to have a phone number.

Phone support is expensive ($8 to $15 per call). Digital support is cheaper ($3 to $8). AI support is cheapest ($0.20 to $2.00). The economics push toward eliminating phone. But phone is also the channel where customers feel most heard, especially for complex or emotional issues. Removing it saves money and degrades experience.

AI is being deployed faster than it's ready. Only 25% of contact centers have successfully integrated AI into their daily operations. The other 75% own AI tools but haven't operationalized them. This creates an awkward middle state: companies reduced human staffing in anticipation of AI handling the load, but the AI isn't actually handling it well. Customers fall through the gap.

Support agents are burned out. HubSpot found that 75% of customer service reps reported their highest-ever ticket volumes in 2024. Zendesk's data shows that 77% of agents say their tasks are more numerous and complex than last year. More than half report feeling burned out. When agents are burned out, quality drops. CSAT declines by 18% when support comes from a high-stress team member. The remaining humans are overwhelmed because AI took the easy tickets, leaving only the hard ones for people who are already stretched thin.

The cost center mentality is winning. The fundamental problem is structural. Customer support is categorized as a cost center. The incentive is to minimize cost. Every dollar saved on support looks good on a P&L. The revenue impact of bad support (churn, negative word-of-mouth, chargebacks) shows up in different line items, often months later, and rarely gets attributed back to the support budget cut that caused it.

This is the most important dynamic and the hardest to fix, because it requires changing how companies think about support at the executive level. It requires connecting support quality to revenue retention, and most companies don't have the data infrastructure to make that connection visible.

The Klarna story

Klarna is the case study everyone cites, and it's worth understanding in full because it captures the entire arc.

In 2024, Klarna cut approximately 700 support jobs and replaced them with an OpenAI-powered chatbot. CEO Sebastian Siemiatkowski announced that the AI handled 2.3 million customer conversations in its first month, doing "the work of 700 employees." It was a massive PR win. Every AI vendor cited it. Every tech blog covered it. It was proof that AI could replace human support at scale.

By mid-2025, the story changed. Customer complaints increased. Satisfaction fell. The CEO admitted: "We went too far." Klarna started rehiring humans. They put software engineers, marketers, and legal staff into call centers as a stopgap, then launched what Siemiatkowski described as an "Uber-type" customer service model: remote, flexible, competitive pay, targeting students and rural workers. The company now says customers should "always have the option to speak with a human."

The pattern is instructive:

1. Company fires support team, replaces with AI
2. AI handles simple queries well, complex queries poorly
3. Complex queries pile up with no humans to handle them
4. Customer satisfaction drops
5. Company realizes the cost savings from AI are offset by churn and complaints
6. Company quietly rehires humans under different job titles

Klarna isn't unique. Forrester found that 55% of employers already regret AI-attributed layoffs, and that over half of those layoffs will be quietly reversed, often with jobs returning offshore or at lower wages. Gartner predicts that by 2027, 50% of companies that cut support staff for AI will rehire for similar functions under different titles.

The pattern is so consistent that it's become predictable. And yet, new companies keep repeating it, because the initial cost savings are real and visible, while the downstream damage takes months to materialize.

What good looks like

I don't want to be purely negative. Some companies are genuinely improving their support through thoughtful use of technology.

The ones doing it well share common traits: they automate simple queries (password resets, order tracking, FAQs) while keeping humans on complex issues. They use AI to assist agents (drafting responses, surfacing context, suggesting actions) rather than replacing them. They measure customer experience alongside cost, and they adjust when the data shows quality declining.

The key difference is intent. Companies that deploy AI to reduce costs while maintaining quality achieve a different outcome than companies that deploy AI to reduce costs, period. The first group invests in classification, routing, quality monitoring, and human escalation paths. The second group deploys a chatbot and turns off the phone number.

What would fix it

I think about this a lot because we're part of the industry. We sell tools that automate support interactions. If the industry is getting worse, we're either part of the problem or part of the solution, and that's a distinction worth being honest about.

The things that would actually fix customer support quality:

Treat support as a revenue function, not a cost center. Support prevents churn. Churn reduces revenue. If companies measured the revenue impact of support quality (and some are starting to), the incentive to cut support would weaken because the cost of cutting would become visible.

Require human access. The CFPB proposed a one-click-to-human rule for financial services in 2024, but that rulemaking has been frozen since the Trump administration effectively shut the agency down in early 2025. The principle is still right, even if the regulation is stalled: if you deploy AI, you must also provide an easy path to a human. Not buried in a menu. Not after 10 minutes with a chatbot. Visible, immediate, one click.

Pay support agents more. The average support agent makes $42,830/year. Agent turnover is 30 to 45%. The cost of turnover ($10,000 to $20,000 per agent) means companies spend enormous sums replacing people they could retain by paying 20% more. Costco pays above-average and has 8% turnover. The math works. Nobody does it because the upfront cost is visible and the turnover savings are distributed.

Measure what matters. Stop optimizing for cost-per-ticket and start measuring customer effort, resolution completeness, and churn correlation. The companies with the best support don't have the lowest cost. They have the best retention. Those are different things.

Use AI selectively. Automate the simple stuff. Keep humans on the hard stuff. Measure constantly. Adjust when the data says quality is declining. This is what we try to build toward with our classification approach: the AI decides what should be automated and what shouldn't, rather than automating everything by default.

Where does this leave us

The support industry is getting worse by measurable, reputable metrics. The decline is driven by cost-cutting, AI deployment that outpaces readiness, agent burnout, and a structural incentive to treat support as a cost to minimize rather than a relationship to invest in.

It's fixable. But fixing it requires companies to care about customer experience as much as they care about margins. And right now, margins are winning.

I'd like to tell you that our company has all the answers. We don't. We sell a classification tool that helps route tickets and automate simple queries. It's good at what it does, but it's one piece of a much larger problem. The larger problem is how the industry thinks about support: as a cost instead of a relationship, as a department to shrink instead of a function to invest in, and as a problem technology can solve instead of a human need that technology can support.

We're going to keep building tools that try to make AI support better rather than just cheaper. Whether the rest of the industry follows is up to the executives making budget decisions. The data suggests they should change course. The question is whether they will.