Customers are giving businesses fewer chances. In a survey of 1,000 U.S. adults released November 9, people said they would put up with an average of 2.78 poor experiences before leaving a company, down from 3.24 in the first quarter of 2026. The number comes from the Q3 2026 AI Consumer Experience Index, run by the research firm Metrigy in partnership with NiCE, which sells AI customer service software.
What did the NiCE and Metrigy Q3 2026 index find?
The index found that customers tolerate fewer bad experiences than they did six months ago (2.78, down from 3.24), that trust is the top thing they want from a business (41%), and that 28.8% would pick an AI agent if they knew their issue would get resolved, up from 23% in Q1.
Metrigy surveys the same-sized panel every quarter and scores eight parts of the service experience against a Q1 2026 baseline of 100. The Q3 wave was fielded August 10 to 14, with a topline margin of error of about 3.1 points, according to the release on Business Wire. The main numbers:
| Finding (Q3 2026, 1,000 U.S. adults) | Number |
|---|---|
| Poor experiences tolerated before leaving | 2.78 (was 3.24 in Q1) |
| Rank trust as the top factor in dealing with a business | 41% |
| Rank resolution as the top factor | 35.4% |
| Would choose an AI agent if resolution is guaranteed | 28.8% (was 23% in Q1) |
| Pick a service channel mainly for accuracy | 54.7% |
| Would pay more for better service | 63.9% |
| Want AI and humans working together, switching by complexity | 39.4% |
Two caveats. NiCE paid for the research and sells the kind of AI it measures. And the 2.78 figure comes with a qualifier in the release: "assuming they could do so." People who can't easily switch, like a customer locked into a contract, will stick around longer than that.
Why are customers giving businesses fewer chances?
The study doesn't say why tolerance dropped. It only measured that it did, by about half a bad experience per customer in two quarters. My read, from running ads since 2013, is that customers now know a faster option is one search or one ad away, so a slow callback costs you sooner than it used to.
Metrigy CEO Robin Gareiss pointed to a newer pressure in the release: consumers are "sending personal AI agents to negotiate, escalate, and decide for them," and those agents "will recommend the brands they can trust." Nobody has measured that yet. But if a customer's AI assistant is picking the plumber, a record of unanswered calls will count against you.
Do customers want to talk to an AI agent now?
More of them do, as long as the AI actually solves the problem. 28.8% would choose an AI agent when resolution is guaranteed, and the most popular setup (39.4%) is AI and humans working together, with the system handing off to a person when the issue gets complicated.
Another 37% want human-led service backed by AI, and 23.6% prefer AI-led service with a person on call. Younger adults lean in hardest. Among 25 to 34 year olds, 74.4% expect AI to improve customer service over the next two years, against 49.5% of everyone surveyed. The most-wanted premium feature, picked by 37%, was being able to choose between an AI agent and a human.
What does this mean for a business that runs ads and answers leads?
It means your first bad experience happens earlier than you think. For a lead from a Facebook ad, the unanswered call, the reply the next afternoon and the booking at the wrong time each count as one. In Q1 the average customer allowed a little over three of those. Now it's under three.
I see this in our own calendar data. Over 45 days, the calls our AI agents booked went 10 showed and 14 no-shows. Most no-shows were booked 2 to 5 days out, and most people who showed were booked within 24 hours. One no-show came down to a wrong time zone on the booking, which is a bad experience the customer never chose. If you want a rough dollar figure for the calls you're missing now, the free missed call calculator will give you one in about a minute.
KenjiAI's Voice AI agents answer inbound calls, qualify the lead and book them on your calendar, then hand the call to a person when it needs one. That's the AI-plus-human setup 39.4% of consumers in this study said they prefer. The texts, reminders and review requests run from the same CRM, so no lead waits until tomorrow for a reply.
What should you check this week?
Go through your own lead flow the way a new customer would and count the bad experiences before anyone helps them. Start with these four checks:
- Call your business line after 6pm and on Saturday. Note who or what answers, and whether the caller can book anything.
- Submit your own ad form and time the first reply. If it takes hours, that alone is one of the 2.78.
- Look at how far out your bookings land. In our data, bookings 2 to 5 days out were where most no-shows came from.
- Make sure every booking shows the customer's own time zone, and that a person can step in when the AI or the front desk gets stuck. For more on that, read what missed calls cost a local business.
Metrigy says the full Q3 index includes breakdowns by persona, industry and channel. If you work in retail, financial services or healthcare, look at your own sector's numbers there before you change anything.