AI Takes Routine Customer Service. Your Team Gets the Hard Cases

What Came Out This Week

CNBC published a story this week about a day in customer service as AI changes the work (CNBC). It follows people on the phones and the companies selling the AI that answers them. Here is what stood out.

Kaylee Harris takes customer service calls for a credit card company. Her shifts used to be a blur of calls, typing notes and reading out long card disclosures (CNBC). These days AI lets her give each customer more of her attention, she told CNBC (CNBC). That is the upside.

Then there is Shannon. She handles customer service for a telecom company in Cardiff, Wales. She asked CNBC to leave out her surname because she worries her employer would figure out who she is (CNBC). The AI notes at her current job are often wrong, she said, so she writes her own (CNBC). At the property services company where she worked before, managers told staff to ask ChatGPT when they did not know an answer (CNBC). AI agents booked appointments, but if a customer's date was not open, the case went to a person (CNBC). Shannon said the team was always putting out fires (CNBC). "All the complicated cases came to us, causing a lot of burnout" (CNBC). That is the downside.

In theory, automation should take the paperwork off agents' plates and leave them more room for the calls that need a person's empathy, judgment and experience (CNBC). That is the promise.

Nikola Mrkšić is CEO and co-founder of PolyAI, a London company that builds voice AI for customer service (CNBC). He said contact-center agents can face "maximum abuse for minimum wage" (CNBC). In his view, AI agents can carry part of that load (CNBC). He said the job is hard on the people who do it. That is the pitch.

Virgin Media O2 is a British telecom and media company. In September it put an AI voice agent to work. The company said the agent takes the routine issues, which leaves human advisors more room for the sensitive or complex cases (CNBC). Debra Maxwell, CEO of the U.K. customer experience company ArvatoConnect, told CNBC she wants technology to handle 70% of interactions and people the other 30% (CNBC). She said a tool ArvatoConnect built for the Crown Commercial Service, the U.K. government's procurement agency, drafts supplier viability assessments at 97% accuracy and saves hours of manual data work (CNBC). That is a real number.

If AI takes the routine work, people may be left with more of the hard cases — and have to cover when the tech breaks (CNBC). That is the part I would underline if I owned the pipeline.

This sits alongside industry reports of rising workloads. According to MaxContact's 2025/26 U.K. contact-center benchmarking report, more than half of the decision-makers surveyed said their agents' workloads had grown, compared with around 40% in 2024 (CNBC). The survey does not say what is behind the increase (CNBC).

Amanda Jones works at King's College London, where she studies how organizations behave and how work is changing. Her specialty is work intensification: jobs that get more demanding over time (CNBC). She told CNBC that call-center work could get more intense if people are left with only the hardest calls (CNBC). She said some customers may already be frustrated before they reach a human (CNBC). She also said that simpler tasks give workers a bit of psychological relief between harder ones, and if those tasks go away, the day can feel much heavier even when the total amount of work stays the same (CNBC). She warned that better output numbers can come with tougher targets, and workers could lose the chance to build the experience they need for more demanding work (CNBC).

Maxwell said handling times may have to go up as staff take on harder cases (CNBC). She said ArvatoConnect judges agents on whether they understood the customer and reached the right answer, not on scripts or time targets (CNBC).

Harris likes what the AI tools do for her, even when glitches add work. But she worries that automating more of the work costs customers something (CNBC). She said the country is chasing speed over quality, and customers are tired of AI being pushed on them (CNBC). Shannon sees good in the tools too, but she does not think AI should ever take the place of real customer service (CNBC). She said, "I believe people buy from people; with AI you don't get to experience that" (CNBC).

My read: if you run a home services company or an RV dealership, you already run a version of this split. The routine questions your team fields all day are the ones AI can take. What is left is lead handling: the caller with a real problem and a real reason to call someone else. My advice: decide which calls are worth a human, put your best people there, and judge them on whether the customer got the right answer, not on how fast the call ended.

Opens salesbeast.com