No.

Well, probably not.

That’s truly the most accurate answer you’re going to get to the question. Nobody knows for sure, but, as of right now, I think there is every reason to believe that us humans, in all the incredible variety of our weird, beautiful, strange bodies, will be gainfully employed to serve customers for decades to come. 

You’re not convinced though, are you? I don’t blame you. It feels like artificial intelligence is in every tool, every article, and every conversation. AI is here, it is real, and it can do remarkable things. 

So let’s get into this more deeply and try to understand how the rise of AI — specifically large language models and generative AI — will shape the future of the work (and the industry) of customer support. To begin with, we should acknowledge the AI-powered, ultra-realistic, simulated elephant in the room: customer service job loss due to AI.

It’s happening now: AI affecting customer service jobs

Person has a job. Person is fired. AI software now does that job.

That is the simplest formulation of the AI-takes-our-jobs scenario, and it does happen. In October 2023, Suumit Shah fired 27 of his customer service staff members and replaced them with a ChatGPT-powered bot which he claimed was “100 times smarter” and “1/100th of the cost” of the existing team.

A cursory Google search shows that Shah’s customers are not universally 100 times happier, so we’ll mark that down as a mixed bag at best.

Shah was at least willing to make his claims out loud. Other company leaders prefer to avoid the press, and they often won’t even admit it internally. Rather than say “we’re replacing people with cheaper/faster AI,” they will use the obscuring and emotionally distancing business language of “restructuring,” “transforming,” and “optimization” to obscure their aims.

CEOs attending Yale SOM’s CEO Summit put customer service right at the top of their transformation list: “When we asked, ‘Which one of these does AI have the greatest potential to transform for my business’, 31% said customer service, while 25% said staffing and workforce–which hardly bodes well for professional services employees.”

Now, if we were feeling generous, we could assume that by transform they mean “add additional technological assistance for our highly valued human service staff,” but the evidence for that is thin at best. 

A safer assumption would be that the future of large scale customer service is smaller human teams with larger AI components — which is not necessarily a problem in the long term, but it’s definitely a concern for anyone needing to earn a living during the transition period. 

Why is customer support the first choice to be “replaced”?

Why not target accountants or marketers or C-suite executives?

Certainly there is a financial incentive. Customer service is a cost center from an accounting perspective, and those costs are very easy to measure right down to a dollars-per-ticket level. The return on those costs, on the other hand, is significantly more complex to track.

Customer service is costly to scale, often requiring 24-hour coverage and additional infrastructure. On a budget sheet, the customer service department must look like a tempting target to a CFO under pressure to improve the numbers.

There are other cost centers in a business: legal, HR, finance. There doesn’t seem to be the same enthusiasm to sweep through those departments with the big broom of AI.

Hard evidence is lacking, but I suspect that many business leaders fundamentally do not understand (or value) customer service as a skill and therefore assume it will be easier to automate.

From the outside, customer service doesn’t seem complicated, particularly if your usual experience as a customer is with under-trained people working through a frustratingly narrow script.

Most people who haven’t spent much time in support have no sense of the breadth and depth it can cover. They don’t know how much nuance and skill goes into high-quality, consistent, reliable support work. One moment you’re answering the most repetitive, obvious question about passwords, the next you’re digging into a sensitive issue with customer data or tracking down a niche browser-related bug, and the whole time you’re detecting and fighting off increasingly convincing social engineering attacks.

Therefore, the “customer service” that executives think they will be replacing, augmenting, or “transforming” with AI tools is a very simple mental model that does not map well to the messy complexity of reality. 

Admittedly, there is fun to have with customer-facing AI that has become very real. Would you buy a Chevy for a dollar? Or perhaps you’d prefer 2,000 chicken nuggets. 

Less fun is a world in which the average level of service quality drops even further, and we’re all stuck playing a sort of infuriating text adventure just to get help.

Right now the explosion of AI support agents and associated businesses suggests there is a lot of experimentation on real customers still to come. When AI bots are so much cheaper than people, many companies will be willing to take the risk. And they are certainly cheaper — at least for now.

But is AI really any cheaper?

Training and running the large language models that enable ChatGPT, Claude, and the other frontrunners in AI is an enormously costly venture, sucking up billions more dollars every year than the owners are able to charge to use them. That equation could change with technological breakthroughs — in January 2025, OpenAI agreed to join other investors in building $500 billion in “AI infrastructure," only to be surprised a week later by the apparently far less costly “DeepSeek”  — but there is no guarantee of a market-changing adjustment in costs.

As consumers — a category which includes every business which relies on an underlying LLM it must rent access to — we do not know what the final costs will be once the gold rush is over and the winners of the market need to make their money back.

Will it truly be cheaper to use an AI to generate high-quality customer service answers than to pay a skilled person to do it? The answer is less clear than the headlines would have you think. 

Certainly there will be types of work which AI-enabled software can automate or completely replace, but we do not yet know the real cost. Smart companies will not want to lose their most skilled employees before they can be certain of how they will replace their work output. Less smart companies, and those that “move quickly and break things,” may find themselves without access to those skills at all.

Humans still need apply

Humans in customer support are more than fleshy search engines connecting customer questions to an existing knowledge base article or canned reply. That is certainly part of the role, but even then a skilled support person will direct the customer not just to the “matching” answer but also point them toward functionality or processes they have not even considered.

What begins as a simple question about a product or service can often become a highly sensitive, complex, and nuanced discussion that may involve coordination with multiple internal teams. Support teams also turn customer questions into insights for sales, marketing, design, and development teams to use.

In addition to their question-answering hats, support people wear enough additional hats of all shapes and sizes to keep a coterie of Royal milliners in business for generations. 

Customer support staff often provide business consulting, billing assistance, technical investigation, and account management. They handle replies from marketing campaigns and bug bounties, write internal and external documentation, respond to hopeful job applicants, and spot social engineering attacks. 

Even after all that, they often find time to share a timely GIF, offer emotional support, appreciate people’s pets, and help their own teammates work through complicated issues.

Much of that work is invisible to people outside the support department, and it’s therefore not on the radar of the non-support people building and selling AI tools. Even if they were aware, applying a giant statistical probability machine to that wide range of problems just isn’t likely to succeed without a lot of technological progress.

Assuming the very best case (best for OpenAI shareholders at least), where AI tools get continually better, only a portion of support work can be sufficiently automated. The more likely “good timeline” outcome, it seems to me, is that AI becomes another assistive tool that can multiply the effort of a smaller number of people to help a larger number of customers. 

If that future is to come to pass, then a consistent, high-quality self-service experience will be the key.

Self-service with AI is the future of support

Self-service technology has a history dating back to at least first-century Egypt, when the first vending machine was built. Having a customer acquire a product, make a transaction, or get assistance without waiting on somebody to help them is both convenient for customers and financially attractive for businesses. 

New technology has always been adapted to self-service usage. In addition to vending machines of every type there are ATMs, transit ticketing machines, airline bag drops, IVR phone systems, and self-service checkouts — all customer service tools making use of automation to improve speed and reduce costs (albeit at varying levels of quality). There is every reason to assume that machine learning and artificial intelligence will be applied more widely as its capabilities grow.

Just like grocery checkouts and bag drops reshaped but did not replace employment, AI will not replace human customer support everywhere. AI may take certain roles, but on the whole it will augment human service. Retail food stores, bank tellers, transit ticket offices, phone support, and grocery store checkout staff all still exist. Technology extended all those forms of service into places and hours that were either impossible or costly to staff, in some cases reducing the number of service staff but rarely replacing them entirely. 

While current AI systems face serious challenges with reliability and accuracy, they are already capable of handling less complex queries, internal tagging, and improved knowledge base searching, at least some of the time.

As those capabilities are rolled out, I expect the first support interaction for most customers, most of the time, will be some form of AI-mediated self-service. 

That could be a chatbot providing relevant knowledge base articles, a phone bot asking basic questions, or an AI-generated email response explaining the most likely answer.

In some businesses, AI may be able to handle a large percentage of incoming support questions. In others, it may only manage a handful of common cases. Either way, the first step in getting support will usually be a form of self-service informed or controlled by AI systems.

Customers will either receive the help they need or pass beyond the capabilities of self-service into the world of person-to-person customer support. Therefore, we should expect that the direct customer support work done by people will tend to be the more complex, nuanced, sensitive, and specific work 

Human roles in the support teams of the near future

The basic structure of an online customer support organization has remained mostly unchanged since the early days of the commercial web: Team leaders and frontline support staff covering multiple channels, all reporting to a head of support.

Even as the mobile web, social media, and chatbots arrived, they were absorbed into the existing structure. Artificial intelligence, on the other hand, promises (or perhaps threatens) to reshape the way we build and staff our teams with people.

Why? First, there is financial pressure exerted by AI. Automating tasks via AI will be cheaper than adding new staff, and AI can amplify the effort of skilled people to help more customers. That could fundamentally shift the ratio of support staff required to help a given number of customers.

Financial impacts aside, the new capabilities and broader usage of AI across industries will mean customers increasingly expect to interact with AI tools. The support teams being built today will certainly include AI as a core element of their structure rather than as an add-on. That will affect the way they hire, train, tag, report, and measure the quality of their service.

All of that, every generative AI system, will need high-quality content to learn from and to point people to. That content must come from people with skill, empathy, and deep knowledge of both the customers and the products and services they use.

Even with all of that content, many scenarios will just require human input. Even if AI tool builders achieve their wildest dreams, which seems increasingly unlikely to happen, people will still be needed to:

  • Manage the AI tools and monitor their output.

  • Create the documentation AI imbibes and shares.

  • Answer the most sensitive, complex, and nuanced problems.

  • Define and shape quality assurance for hybrid AI/human support.

  • Run classes and events to serve customers at scale.

  • Manage accounts and build individual relationships.

  • Use AI tools to accelerate queue management, for reporting, and to prioritize tasks.

  • Extract insights from customer interactions. 

Whichever element of customer support you are personally most interested in or most suited to, I am confident that there will remain support roles suited to you. Those roles may be fewer in number, and it may be that you need to take some active steps to determine where you end up.

Ready to get started? Grab your compass, fill your water bottle, and meet me over at Navigating a Support Career in an AI-Powered World.