April 2, 2026
BiancoBlue/Depositphotos.com
Can AI Customer Service Actually Replace Humans?
In a report for CNBC, journalist Kevin Williams outlined the current struggles that AI customer service bots and call centers are having in satisfying shopper demands, despite several industries adopting these solutions at scale.
Noting that while the promise of AI call centers and chatbots is enticing for many operations looking to slash costs, deflection — rather than resolution — is often a part of the game as it stands, with customer satisfaction often being the metric left behind.
“I hate AI customer service chatbots,” said Californian shopper Carmen Smith, as quoted by Williams.
“It seems that no matter what, they all will either point you to some type of FAQs list or repeat information you’ve already tried and found lacking,” Smith said. “I hate dealing with them, but a lot of companies use them nowadays, alas. I’d rather speak to a human being.”
Qualtrics data cited by the report indicated that ~20% of consumers who had engaged with AI customer service agents received zero benefit from the interaction, a failure rate four times higher than AI usage more broadly.
“Too many companies are deploying AI to cut costs, not solve problems, and customers can tell the difference,” said Isabelle Zdatny of Qualtrics, author of the 2026 Customer Experience Trends Report.
Other notable trends captured by Williams:
- AI systems often reflect leadership intent in terms of refunds, reduction of escalation, and call times. While human agents may exert some degree of perceived compassion or rapport with customers, ultimately an argument can be made that they, too, are simply scripted.
- A frequent complaint regarding AI customer service centers or solutions is that they make shoppers feel as if they trapped in a loop via automation. This presents a frustrating, annoying (or worse) barrier.
- AI customer service may become the de facto standard: Zendesk CEO Tom Eggemeier underscored his belief that within three years, half of all online customer service interactions will be handled by AI bots, with this number rising to 80% within the next five years.
If AI Customer Service Agents Are the Future, Proponents Say There Are Upsides (But Do Consumers Agree?)
On the other hand, some proponents argue that AI deflection can be valuable to protect high-burnout workers, mitigate turnover, and reduce mental health strains. Further, they may be able to simply deliver the necessary “no” that a human may struggle with or hedge against.
“If two people are arguing about a refund and the law says it is not available, a judge would adjudicate rather than argue continuously back and forth. That is often what happens in agent-to-unhappy-customer scenarios,” Terra Higginson, principal research director at Info-Tech Research Group, said.
“This makes the process about enforcing rules and regulations rather than making refunds difficult… without the arguing and back-and-forth strain of being yelled at for following company rules,” she added, pivoting to also highlight that obstructing valid refunds was not the goal of an ethical or effective AI agent.
Another recent survey seems to indicate that consumers aren’t as in love with AI customer service as those deploying them. According to CX Dive, which pulled figures from a 2025 HubSpot and SurveyMonkey survey capturing the thoughts of 1,800 business leaders and 15,000 consumers worldwide, a majority (53%) of shoppers “actively dislike or hate AI’s use in service interactions.”
“This widespread aversion suggests that current AI applications often fail to meet customer expectations for empathy, nuance and genuine connection,” said Zoe Padgett, senior research scientist at SurveyMonkey.
“Rather than seeing AI as a helpful tool, many customers perceive it as a barrier, leading to frustration and a feeling that their unique concerns aren’t truly understood.”
BrainTrust
Alex Siskos
SVP Strategy, Everseen
John Hennessy
Retail and Brand Technology Tailor
Mohamed Amer, PhD
Strategy Advisor, CEO & Co-Founder, BridgeCommAI
Recent Discussions
Discussion Questions
If, and when, AI replaces human customer service, will it be deemed a success? Do you foresee this occurring over the course of the next five years?
Concerning retail customer service and the integration of AI chatbots, what are your broad feelings about its current deployment? What opportunities to improve exist?
Which customer service aspects are humans best suited for? What might AI excel in?
Poll
Loading …
The key to any AI chatbot adoption is ensuring you have the right exit points for the customer when the AI is not delivering what they need. They fail when customers get stuck in the non-answer loop.
The best chatbots offer full self service and give an exit ramp(s) to escalate to a human in the lead when they can no longer provide the next step.
This is by no means easy because it requires connected systems, thoughtful design and the right staffing levels to make it appear seamless to the consumer.
I don’t think that word “seamless” is a descriptor customers would use!
Broad ai search can pull up a retailer’s FAQ answers faster than ever.
So the simple minded chatbots are themselves being replaced.
Retail site chatbots will be asked the same difficult questions that Twitter CS Agents are asked. Order issues or site functions… personalized issues that Twitter reps push into direct messaging. If the site chatbot is trained on Twitter questions, it would have to be trained on the DM’d answers to those questions.
When online search results overwhelm or miss the mark, AI-driven chatbots can refine the parameters for faster, relevant results.
A chatbot can surface suitable products like spring dresses. However, when I updated a prompt to say, “Please exclude ugly ones,” the chatbot promptly ended the conversation.
Every one of us has been trapped in “voicemail jail” where the system is simply not capable of helping. The system I chatted with at Citibank the other evening had me yelping in frustration. Companies can begin calls with AI, but should give customers the option of upgrading to chatting with a human.
I agree (and not just re: Citibank!).
It begs the question of why the company set up the chatbot (ai or not).
Was UX the primary goal (happy customers buy more & come back)?
Or, once you’re in chatbot jail, do you suspect the primary reason was to let go of the humans in the jobs.
“Trapped in a loop due to automation” were words spoken to me from a CIO at the Dallas AI Convergence earlier this week when talking about AI replacing human customer service. And at the core of his statement was not “when and if” but rather “how and why” do I automate parts of the customer service offering – especially with what is evolving at a crazy pace with AI chatbots/agents. And what do we base that off of? Your last question – humans are better in empathy and problem solving. And that will always be the baseline measurement. Moving beyond the easy FAQs, and more into the ability for an agent to reason and decide when to bring in the human in the loop is what will define success for me in the future. I see this as iterative and those who will win here sooner that others, are those that will focus their investment of time and resources to continuously refine the models in a domain specific arena.
Why must it be an “either-or” condition? In most call centers there are levels of agents because it’s just not cost effective to have highly experienced, Level 4 agents answer every call. And let’s not forget that before we were stuck in technical loops we were stuck in “inexperienced Level 1 rep loops” and then in “outsourced, inexperienced level 1 loops” and then in “chatbot loops” and then “NLP Chatbot Loops.” The technology may have changed but the pain remained the same.
This is not necessarily a “how” problem; it’s a “what” problem and regardless of the “how” (AI or human or other), companies must own, monitor and control the customer experience if they wish to protect their brands.
Bradley Cooper’s answer is spot on. I believe AI is definitely capable of handling repetitive, easy to answer questions with great ease (Level 0 or 1). And that means greater efficiencies and customer service for 50-60% of all calls (maybe more). But let’s use Bradley’s exit ramps and staff them with humans.
Q: As we are decades into digital as primary shopping channel… aren’t the # of Level 0 or 1 questions to a retail chatbots dropping significantly?
Non proprietary search (google ai) can answer the simplest… eg how much is a return shipping label? What address do I ship to to return on my own? Is policy 30 days from receipt or 30 days from order date.
The customer frustration with chatbot loops is real, as the personalization (level of chatbot, or a FAST offramp to human) are still lacking.
These are design & programming issues. Not the shiny object of how may acres the newest AI data center will be.
No. Chatbots are NOT able to replace customer service by humans. That won’t stop far too many retailers from doing it. But the downsides are legion. They are slow, frustrating for customers, lack the natural elasticity needed to keep customers happy (read Street Level Bureaucracy), and destroy brand goodwill as a result.
AI can replace humans in customer service the moment it can consistently do seven things well:
- Diagnose the real problem (not just the stated one)
- Recommend solutions with a high probability of success
- Access the right systems and take action within them
- Apply the right level of recovery when the company is at fault
- Validate claims without creating friction (guarding against fraud/abuse)
- Anticipate subsequent downstream issues before they happen
- And, most importantly, leave the customer feeling understood and valued
We’re not there yet but progress is being made on the first six and there is no reason to doubt AI’s ability to get there within the next 5 years on those. The seventh is the real test.
Exactly. Isn’t that the promise of all AI?
At first customer service seems a natural for AI. The least common issue would be as easy to surface as the most common. No variance in training.
The reality is far from that simple.
Whether it’s due to poor data, bad training or other issues or all these issues, AI customer service is poor. We did a deep dive on a couple examples and it wasn’t pretty. Sometimes guardrails which I assume were put in to get a human review prevented further transactions. The customer was effectively locked out for a time by engaging with the AI.
Much promise. Much work to be done.
If “success” is defined by corporate balance sheets (slashing labor costs), then yes, it’s already succeeding. But if success is defined by customer resolution, we are a ways off. In five years, we likely won’t see total replacement, but rather “aggressive tiering.” You’ll only get a human if you are a high-value loyalty member or if your problem is expensive enough to merit a person’s time. The friction we feel today is because companies are trying to force AI to do human type tasks (handling nuance) to save a buck. Until AI is used to empower customers with utility rather than just deflect and stall them, the dislike stats won’t budge.
You’ve identified the “follow the money” route!!
Ultimately, the consumer will decide whether AI bots predominate in retail. However, the likely outcome is mixed. AI can do some things well – mostly simple tasks such as providing information, giving basic updates, aiding search, and so forth. It struggles more with complex and involved tasks; although this may change in the future. What it cannot do is nuance, emotion and connection – all very human traits that customers value highly.
Did ATMs replace bank tellers? (Did video kill the radio star?) I boldly predict that AI will not 100% replace humans in the next five years. My annual customer service/CX research shows 7 out of 10 consumers still prefer the phone to resolve problems and deal with complaints. Related to AI 65% of customers expect AI to become the primary mode of customer service in the future (up from 63% in 2025), however a human is still expected to be in the loop when needed.
An AI-first, human-second strategy works, where a company pushes its customers to use self-service solutions, however, without easy access to a live agent, the company/retailer becomes a commodity.
(For a free copy of the research – and you don’t have to share your email address to get it – go to https://www.Hyken.com/research.)
Context. If/Then. How quickly in the ‘conversation’ can AI get to the real problem and a range of solutions? Can AI be taught enough nuanced If/Then scenarios that it can solve the problem without exhausting the customers patience? There’s very little in today’s chatbox experiences that would help predict AI success in customer service. But then again, we are probably still at the grade school of AI deployment. Let’s see what the experience feels like when AI hits high school and college. Timing? Who knows.
AI-driven customer service will only be deemed a success if retailers can demonstrate clear, data-driven improvements — not just cost savings. Metrics such as resolution rates, customer satisfaction, sentiment analysis, and repeat contact frequency should guide adoption. Without both objective and subjective measures, it’s difficult to determine whether AI is improving service or simply creating new friction for customers.
Today, AI chatbots are most effective when handling routine, high-volume inquiries such as order status, returns, store hours, and basic product questions. Where AI often falls short is in complex, emotional, or exception-based scenarios, where human judgment, empathy, and flexibility remain critical. The most successful implementations are those that augment human teams with clear escalation paths rather than attempting to fully replace them.
Over the next five years, AI will likely play a much larger role in customer service, but full replacement of humans seems unlikely. AI excels at speed, consistency, and scale, while humans remain best suited for nuanced problem-solving and relationship-building. Retailers that strike the right balance — and measure performance carefully — will improve both efficiency and customer experience, while those focused solely on cost reduction risk undermining customer trust and loyalty.
True up until 2026:
“ AI chatbots are most effective when handling routine, high-volume inquiries such as order status, returns, store hours, and basic product questions.”
AI chatbots are just a small portion of “AI”
Those most low level Q&A, have already been made redundant.
A search page, uses AI to answer the same Q’s without the customer having to go to any retailer’s proprietary chat channels (site, Twitter, FB).
When a customer does show up on the retailer’s proprietary channels…becoming likely questions asked will continue to be higher level from the start.
These are chatbot end-goal & design issues. On an everyday level, should the chatbot be able to correctly & swiftly handle mid to upper middle customer scenarios?
Before AI, we were writing on these pages about what was wrong with human customer service. I think we found similar problems.
AI promises much, and from what I have seen and used, it is on the way. But when it comes to customer service, I want the same from the human as from the AI.
I call. They ask, “How can I help you?” They immediately send me to the place that solves my problem. The operative word is “immediately”. Am I asking too much? I see AI performing better than humans in this scenario, eventually.
Can AI customer service actually replace humans? Short answer: Nope.
AI can support a human service agent, but by on its own it’s annoying. Too many buttons to press, too many times repeating what you just said. Too many wrong answers. I’ll take human help over AI any time.
AI customer service bots will replace a large portion of routine interactions, and this will reshape how service is delivered in retail. They are effective at handling simple queries, prioritizing customer requests, and resolving minor issues quickly. This improves efficiency for retailers and reduces wait times for customers.
However, relying only on AI can create frustration when issues require judgment or deeper understanding. Customers still expect human support when something goes wrong. The focus should be on using AI to handle volume while ensuring clear and quick access to human assistance when needed. Retailers that strike this balance will improve service efficiency without compromising customer trust.
More Discussions
Will Automated Pay Stations Pay Off For Costco?
Costco is making a play into improving its checkout process.Is Bed Bath & Beyond’s Purchase of The Container Store a Smart Play?
Was the acquisition a good move?Are NYC’s Laws Hampering Instacart’s Delivery Model?
Are NYC’s laws hurting business for Instacart, and its customers?
The key to any AI chatbot adoption is ensuring you have the right exit points for the customer when the AI is not delivering what they need. They fail when customers get stuck in the non-answer loop.
The best chatbots offer full self service and give an exit ramp(s) to escalate to a human in the lead when they can no longer provide the next step.
This is by no means easy because it requires connected systems, thoughtful design and the right staffing levels to make it appear seamless to the consumer.
I don’t think that word “seamless” is a descriptor customers would use!
Broad ai search can pull up a retailer’s FAQ answers faster than ever.
So the simple minded chatbots are themselves being replaced.
Retail site chatbots will be asked the same difficult questions that Twitter CS Agents are asked. Order issues or site functions… personalized issues that Twitter reps push into direct messaging. If the site chatbot is trained on Twitter questions, it would have to be trained on the DM’d answers to those questions.
When online search results overwhelm or miss the mark, AI-driven chatbots can refine the parameters for faster, relevant results.
A chatbot can surface suitable products like spring dresses. However, when I updated a prompt to say, “Please exclude ugly ones,” the chatbot promptly ended the conversation.
Every one of us has been trapped in “voicemail jail” where the system is simply not capable of helping. The system I chatted with at Citibank the other evening had me yelping in frustration. Companies can begin calls with AI, but should give customers the option of upgrading to chatting with a human.
I agree (and not just re: Citibank!).
It begs the question of why the company set up the chatbot (ai or not).
Was UX the primary goal (happy customers buy more & come back)?
Or, once you’re in chatbot jail, do you suspect the primary reason was to let go of the humans in the jobs.
“Trapped in a loop due to automation” were words spoken to me from a CIO at the Dallas AI Convergence earlier this week when talking about AI replacing human customer service. And at the core of his statement was not “when and if” but rather “how and why” do I automate parts of the customer service offering – especially with what is evolving at a crazy pace with AI chatbots/agents. And what do we base that off of? Your last question – humans are better in empathy and problem solving. And that will always be the baseline measurement. Moving beyond the easy FAQs, and more into the ability for an agent to reason and decide when to bring in the human in the loop is what will define success for me in the future. I see this as iterative and those who will win here sooner that others, are those that will focus their investment of time and resources to continuously refine the models in a domain specific arena.
Why must it be an “either-or” condition? In most call centers there are levels of agents because it’s just not cost effective to have highly experienced, Level 4 agents answer every call. And let’s not forget that before we were stuck in technical loops we were stuck in “inexperienced Level 1 rep loops” and then in “outsourced, inexperienced level 1 loops” and then in “chatbot loops” and then “NLP Chatbot Loops.” The technology may have changed but the pain remained the same.
This is not necessarily a “how” problem; it’s a “what” problem and regardless of the “how” (AI or human or other), companies must own, monitor and control the customer experience if they wish to protect their brands.
Bradley Cooper’s answer is spot on. I believe AI is definitely capable of handling repetitive, easy to answer questions with great ease (Level 0 or 1). And that means greater efficiencies and customer service for 50-60% of all calls (maybe more). But let’s use Bradley’s exit ramps and staff them with humans.
Q: As we are decades into digital as primary shopping channel… aren’t the # of Level 0 or 1 questions to a retail chatbots dropping significantly?
Non proprietary search (google ai) can answer the simplest… eg how much is a return shipping label? What address do I ship to to return on my own? Is policy 30 days from receipt or 30 days from order date.
The customer frustration with chatbot loops is real, as the personalization (level of chatbot, or a FAST offramp to human) are still lacking.
These are design & programming issues. Not the shiny object of how may acres the newest AI data center will be.
No. Chatbots are NOT able to replace customer service by humans. That won’t stop far too many retailers from doing it. But the downsides are legion. They are slow, frustrating for customers, lack the natural elasticity needed to keep customers happy (read Street Level Bureaucracy), and destroy brand goodwill as a result.
AI can replace humans in customer service the moment it can consistently do seven things well:
We’re not there yet but progress is being made on the first six and there is no reason to doubt AI’s ability to get there within the next 5 years on those. The seventh is the real test.
Exactly. Isn’t that the promise of all AI?
At first customer service seems a natural for AI. The least common issue would be as easy to surface as the most common. No variance in training.
The reality is far from that simple.
Whether it’s due to poor data, bad training or other issues or all these issues, AI customer service is poor. We did a deep dive on a couple examples and it wasn’t pretty. Sometimes guardrails which I assume were put in to get a human review prevented further transactions. The customer was effectively locked out for a time by engaging with the AI.
Much promise. Much work to be done.
If “success” is defined by corporate balance sheets (slashing labor costs), then yes, it’s already succeeding. But if success is defined by customer resolution, we are a ways off. In five years, we likely won’t see total replacement, but rather “aggressive tiering.” You’ll only get a human if you are a high-value loyalty member or if your problem is expensive enough to merit a person’s time. The friction we feel today is because companies are trying to force AI to do human type tasks (handling nuance) to save a buck. Until AI is used to empower customers with utility rather than just deflect and stall them, the dislike stats won’t budge.
You’ve identified the “follow the money” route!!
Ultimately, the consumer will decide whether AI bots predominate in retail. However, the likely outcome is mixed. AI can do some things well – mostly simple tasks such as providing information, giving basic updates, aiding search, and so forth. It struggles more with complex and involved tasks; although this may change in the future. What it cannot do is nuance, emotion and connection – all very human traits that customers value highly.
Did ATMs replace bank tellers? (Did video kill the radio star?) I boldly predict that AI will not 100% replace humans in the next five years. My annual customer service/CX research shows 7 out of 10 consumers still prefer the phone to resolve problems and deal with complaints. Related to AI 65% of customers expect AI to become the primary mode of customer service in the future (up from 63% in 2025), however a human is still expected to be in the loop when needed.
An AI-first, human-second strategy works, where a company pushes its customers to use self-service solutions, however, without easy access to a live agent, the company/retailer becomes a commodity.
(For a free copy of the research – and you don’t have to share your email address to get it – go to https://www.Hyken.com/research.)
Context. If/Then. How quickly in the ‘conversation’ can AI get to the real problem and a range of solutions? Can AI be taught enough nuanced If/Then scenarios that it can solve the problem without exhausting the customers patience? There’s very little in today’s chatbox experiences that would help predict AI success in customer service. But then again, we are probably still at the grade school of AI deployment. Let’s see what the experience feels like when AI hits high school and college. Timing? Who knows.
AI-driven customer service will only be deemed a success if retailers can demonstrate clear, data-driven improvements — not just cost savings. Metrics such as resolution rates, customer satisfaction, sentiment analysis, and repeat contact frequency should guide adoption. Without both objective and subjective measures, it’s difficult to determine whether AI is improving service or simply creating new friction for customers.
Today, AI chatbots are most effective when handling routine, high-volume inquiries such as order status, returns, store hours, and basic product questions. Where AI often falls short is in complex, emotional, or exception-based scenarios, where human judgment, empathy, and flexibility remain critical. The most successful implementations are those that augment human teams with clear escalation paths rather than attempting to fully replace them.
Over the next five years, AI will likely play a much larger role in customer service, but full replacement of humans seems unlikely. AI excels at speed, consistency, and scale, while humans remain best suited for nuanced problem-solving and relationship-building. Retailers that strike the right balance — and measure performance carefully — will improve both efficiency and customer experience, while those focused solely on cost reduction risk undermining customer trust and loyalty.
True up until 2026:
“ AI chatbots are most effective when handling routine, high-volume inquiries such as order status, returns, store hours, and basic product questions.”
AI chatbots are just a small portion of “AI”
Those most low level Q&A, have already been made redundant.
A search page, uses AI to answer the same Q’s without the customer having to go to any retailer’s proprietary chat channels (site, Twitter, FB).
When a customer does show up on the retailer’s proprietary channels…becoming likely questions asked will continue to be higher level from the start.
These are chatbot end-goal & design issues. On an everyday level, should the chatbot be able to correctly & swiftly handle mid to upper middle customer scenarios?
Before AI, we were writing on these pages about what was wrong with human customer service. I think we found similar problems.
AI promises much, and from what I have seen and used, it is on the way. But when it comes to customer service, I want the same from the human as from the AI.
I call. They ask, “How can I help you?” They immediately send me to the place that solves my problem. The operative word is “immediately”. Am I asking too much? I see AI performing better than humans in this scenario, eventually.
Can AI customer service actually replace humans? Short answer: Nope.
AI can support a human service agent, but by on its own it’s annoying. Too many buttons to press, too many times repeating what you just said. Too many wrong answers. I’ll take human help over AI any time.
AI customer service bots will replace a large portion of routine interactions, and this will reshape how service is delivered in retail. They are effective at handling simple queries, prioritizing customer requests, and resolving minor issues quickly. This improves efficiency for retailers and reduces wait times for customers.
However, relying only on AI can create frustration when issues require judgment or deeper understanding. Customers still expect human support when something goes wrong. The focus should be on using AI to handle volume while ensuring clear and quick access to human assistance when needed. Retailers that strike this balance will improve service efficiency without compromising customer trust.