Skip to content
ScaleAlphaSi
All posts

AI Customer Service Statistics: What the Research Shows

AI customer service statistics from Pew, the CFPB, NBER and a call center study: public views of AI, where chatbots fail and how long callers wait.

11 min readPrathik KandimallaUpdated

Photo: MARELBU, CC BY 3.0, edited

A glowing orange headset chip feeding rising chart bars on a dark circuit board

AI customer service statistics show public doubt and gains for support staff. Pew Research Center (opens in a new tab) found 41% of U.S. adults would feel worse about a service experience if they later learned they had talked with an AI chatbot.

The same Pew survey (opens in a new tab) found a larger share, 53%, said their view would not change. The warning is real but limited, so any agent you add has to make reaching a person quick and easy.

The Customer Agent is built around that hand-off. It answers calls, texts and web chats day or night. It books service visits, gives status updates from the job record and captures after-hours requests for your morning queue.

Upset customers go straight to a named person with the full thread, and the agent stops replying. Questions about quotes or prices go to the Sales Agent.

For setup steps, read AI Customer Service for Small Business: What to Hand Off. For adoption figures, see Small Business AI Statistics: What the Surveys Show.

Where do AI customer service statistics come from?

These statistics come from a federal consumer agency, the Federal Reserve Banks, a peer-reviewed call center study and two nonprofit research groups. Each counted a different group, from U.S. adults to support workers, bank callers and small employer firms. The table names the group and year behind each headline figure.

Headline figureWho was countedSource and year
41% would feel worse about a service experience after learning it was an AI chatbot; 53% no changeU.S. adultsPew Research Center survey, 2025 (opens in a new tab)
50% more concerned than excited about AI in daily lifeU.S. adultsPew Research Center report, 2025 (opens in a new tab)
Issues resolved per hour rose 14% on average5,179 customer support workers, while an AI assistant was rolled out in stagesNBER working paper, 2023 (opens in a new tab)
About 37% of the U.S. population interacted with a bank's chatbot in 2022U.S. population, a market research estimate the CFPB citesCFPB report, 2023 (opens in a new tab)
About 20% of callers who asked for a person hung up before being servedCallers to one bank's call center, 1999Wharton call center study, 2005 (opens in a new tab)
Accuracy topped AI users' challenges at 46%Small employer firms that use AI, convenience sampleSmall Business Credit Survey, 2025 (opens in a new tab)
98% own a cellphone, 91% a smartphoneU.S. adultsPew Research Center fact sheet, 2026 (opens in a new tab)

Read each row with its group in mind. A national poll of adults, a study of one bank's phone line and a survey of small firms answer different questions, so compare figures only within the same kind of group.

How do customers feel about AI in service?

Customers are wary of AI in daily life, yet most would let it help at least a little with their own day-to-day tasks, according to a Pew Research Center survey of U.S. adults. Service help drew a milder reaction than AI in politics or news. That leaves room for an agent if the service holds up.

Pew Research Center's report on how Americans view AI (opens in a new tab) found 50% of U.S. adults are more concerned than excited about the increased use of AI in daily life, up from 37% in 2021.

Expect many customers to arrive doubtful, so the first answer they get has to be correct and useful.

The same report (opens in a new tab) found nearly three-quarters would let AI assist at least a little with day-to-day tasks and activities.

Pew asked about people's own tasks rather than a business's agent. Routine bookings and status checks are the closest fit, while judgment calls stay with your people.

Pew's survey on reactions to AI involvement (opens in a new tab) asked 5,023 U.S. adults in June 2025 about seven situations, from news articles to customer service.

A national random sample, weighted to match U.S. adults, makes the finding a fair read of how they feel, though your own customers may differ.

In the same survey (opens in a new tab), the most negative reaction came for a political candidate who used AI to help write a speech: 71% would like the candidate less.

The risk to an owner is real, but it sits near the low end of what Pew measured.

What does AI do for support teams?

Support staff given an AI assistant resolved more issues per hour in a large study of customer support work, and newer workers gained the most. The study measured people taking AI suggestions as they worked. It did not test software answering customers on its own, so read it as evidence about assisted staff.

An NBER paper titled Generative AI at Work (opens in a new tab) followed 5,179 customer support workers as a generative AI conversational assistant was introduced in stages.

The paper (opens in a new tab) found access to the assistant raised issues resolved per hour by 14% on average, including a 34% improvement for novice and low-skilled workers.

For a small shop, that kind of gain can matter most where experience is thin, such as a new hire covering the phones.

The authors (opens in a new tab) report little change for experienced and highly skilled workers, and they offer suggestive evidence that the assistant spread the practices of more able workers.

If that holds for a small team, writing down how your best person answers is the first job.

The same paper (opens in a new tab) also finds that AI assistance improves customer sentiment and increases employee retention. Treat those results as a direction for your own shop rather than a promise.

Where do chatbots fall short?

Chatbots struggle once a problem turns complex, according to a Consumer Financial Protection Bureau report on consumer finance (opens in a new tab). The report describes customers caught in loops of scripted replies with no way to reach a person. It also lists wrong answers, wasted time and frustration among the harms customers described.

The CFPB's report on chatbots in consumer finance (opens in a new tab) cites a market research estimate that about 37% of the U.S. population interacted with a bank's chatbot in 2022.

Many of your customers may have used a bank's chatbot already, and they can judge yours against it.

The same report (opens in a new tab) found that each of the 10 largest commercial banks had deployed chatbots as part of their customer service.

Chatbots are common at large banks, so if you add one, the useful question is how to set it up well.

The report (opens in a new tab) says chatbots may help with basic questions, but their usefulness fades as problems grow complex.

Those findings come from consumer finance, yet the lesson carries over to any shop. Give the agent narrow, routine jobs and a short path to a person.

The Customer Agent answers common questions from your own approved answers and job records. It sends angry or legal messages to a named person with the whole thread, then stops replying.

If your customers wait too long for a plain answer, book a free consult meeting. We will go through your hold times and after-hours messages with you and show which routine requests the Customer Agent can take.

How long do callers stay on hold?

Callers at one bank's call center would wait 446 seconds on average before hanging up, by a model-based estimate in a call center study published in the Journal of the American Statistical Association (opens in a new tab). The same study (opens in a new tab) found about 20% of callers who asked for a person hung up before being served.

A man seen from behind holding a black corded handset to his ear beside a lilac wall
A man on a corded phone, back to the camera.Photo: Felicity Tai, Pexels

The study (opens in a new tab) covered all 12 months of 1999 at a small call center for one of Israel's banks, call by call.

It is one older study of one bank, so read its numbers as a picture of caller patience rather than a forecast for your own line.

The same paper (opens in a new tab) found 65% of the bank's customers completed their service in the automated phone menu, and the other 35% asked for a person.

Self-service handled most callers there, yet a large group still wanted a person, so plan the hand-off as carefully as the answers.

For regular service calls, the authors (opens in a new tab) also saw a spike in hang-ups a few seconds into the queue, and another at about 60 seconds, when the recorded wait message played again.

A recorded message may not hold callers the way it is meant to, and the authors leave that question open.

David Maister's paper on queue psychology (opens in a new tab) argues that a wait with no known end feels longer than one of known length, and that an unexplained wait feels longer than an explained one.

A status text that says the part is ordered, or that the tech is on the way, gives the customer an explained wait instead of a silent one.

The Customer Agent sends those updates from the job record in wording you approve, and it never makes up an arrival time.

What do the figures mean for a small business?

The figures point to a narrow, careful start. Most adults would let AI help at least a little with daily tasks, assisted staff gain most where experience is thin, and chatbots fail people when no person is within reach. So give an agent the repeat requests first, and keep refunds, disputes and upset customers with your team.

Among small employer firms that use AI, the Federal Reserve Banks' Small Business Credit Survey (opens in a new tab) found accuracy was the top challenge, named by 46%.

Its report (opens in a new tab) describes the survey as a convenience sample of small employer firms with 1 to 499 employees.

An agent that answers only from your approved answers and job records, with a person reading its early replies, speaks to that worry.

Pew Research Center's mobile fact sheet (opens in a new tab) reports that 98% of U.S. adults own a cellphone of some kind, and about nine-in-ten (91%) own a smartphone.

Nearly every adult owns a phone that can get a booking confirmation or a status update by text.

Turn the figures into a first month:

  1. Count the waits. Note hold times, after-hours messages and how often callers ask for a person.
  2. Pick the routine requests. Start the agent on status questions and bookings. Add reschedules once its answers hold up.
  3. Name the hand-off. Decide who gets upset customers, refunds and disputes, with the full thread.
  4. Check the answers. Read a sample of replies each week until they hold up.

Example: a garage door company sorts its morning calls

Evander runs a garage door repair company with three techs and one office manager. Each morning, calls stack up on hold while the office manager books the day's jobs.

Most of those calls ask the same things. Is the spring in yet, when is the tech coming, can the visit move to Friday?

He starts the Customer Agent on the status and booking calls and texts. It answers status questions from the job record, books service visits into open slots and takes after-hours requests for the morning queue.

Questions about the price of a new door go to the Sales Agent, which drafts quotes from his price list.

An upset caller hears that a person will call back today. The office manager gets the full thread, and the agent stops replying.

For the first few weeks, the office manager reads a sample of replies against the job records. Once the answers hold up, Evander lets the agent handle reschedules too.

Frequently asked questions

Do customers mind dealing with an AI agent?

Customers differ on it. In Pew's survey (opens in a new tab), a sizable minority would feel worse about a service experience after learning a chatbot handled it, while a larger share said it would change nothing.

For an owner, the safe course is quick, accurate answers and an easy route to a named person.

Does AI make support staff more productive?

Support staff can be more productive with AI help. In the NBER study of customer support workers (opens in a new tab), access to an AI assistant raised issues resolved per hour, most of all for newer workers.

The study tracked people using AI suggestions, so it says little about software answering customers alone.

Which of these figures fit a small shop?

The customer surveys fit best, because they asked U.S. adults like the people who call your shop.

The chatbot and call center figures come from banks. Use them as warnings about hand-offs and hold times rather than as targets.

What should I measure before adding an agent?

Measure the waits your customers feel today. For a few weeks, count calls that ring out or sit on hold, after-hours messages and how often callers ask for a person.

Then compare the same counts once the agent is live, so you judge it on your own numbers.

Where can I find small business AI adoption figures?

Our guide Small Business AI Statistics: What the Surveys Show collects federal and research survey figures on how many small firms use AI, which tasks they start with and what holds owners back.

Next step

Book a free consult meeting and bring a rough count of last week's calls, texts and after-hours messages. We will show where the Customer Agent fits and who on your team takes each hand-off.

If the fit is there, you can try any agent free for 30 days.

Our standards

About the ScaleAlphaSi team

ScaleAlphaSi is a Kansas City AI advisory firm specializing in Advanced Marketing and AI Engineering. Our team drives demand through SEO, AEO, GEO and paid ads management. Once that demand arrives, we build specialized AI agents to scale operations for profitability. Each guide draws on engagements with owners and operations teams, covering where to start, where people stay in charge and how to scale one workflow at a time.

Meet the Team
Free consultRequest a Call

Free consult meeting · 30-day free AI agent demo

Book your free consult meeting

Share how your business operates today. We will identify where targeted marketing or AI agents can reduce manual work, lower operating costs or improve margins, then prioritize the highest-impact changes.