Can AI replace a customer service agent?

Reviewed by Colin Balfe on Oct 27, 2026 · Updated Oct 27, 2026 · Graded on a five-step evidence ladder
Partly
The answer

Partly. AI assistants now handle a large share of routine customer chats at big companies, but firms that cut deepest have brought humans back, and AI agents have given wrong answers or been talked into harmful actions.

First shown: Feb 27, 2024. Klarna said its OpenAI-powered assistant handled 2.3 million conversations in its first month, two-thirds of its customer service chats, doing the work of about 700 full-time agents. (VKTR (Klarna release))

What it means

Customer service chatbots used to follow fixed scripts. Large language models can instead read a question, look up account details and reply in natural language in many languages. In February 2024 payments company Klarna said its new AI assistant handled two-thirds of its customer chats in its first month, cut resolution times from 11 minutes to under 2, and matched human agents on satisfaction, by its own figures.

Others followed. Salesforce's chief executive said in September 2025 that AI agents were handling half of its customer service interactions, as reports put its support staff cut from about 9,000 to about 5,000. A September 2025 Gartner survey found 11% of companies had cut staff because of AI and 22% had stopped replacing staff who left.

Then came the walk-backs. In May 2025 Klarna's chief executive said cost had been 'a too predominant evaluation factor', the result was 'lower quality', and the company was hiring human agents again, promising customers there would 'always be a human if you want'. In August 2025 Commonwealth Bank of Australia reversed 45 job cuts tied to an AI voice bot, admitting an 'error' after its union said call volumes were rising. Gartner predicts half of the companies that planned big cuts to service staff will abandon those plans by 2027.

Mistakes carry real costs. In 2024 a Canadian tribunal held Air Canada liable for its chatbot's wrong advice about bereavement fares, rejecting the idea that the bot was responsible for itself. In 2026 attackers persuaded Meta's AI support agent to attach new email addresses to high-profile Instagram accounts and took them over.

The evidence ladder

Every capability on this site is graded on the same five steps, from a single demonstration to real-world consequences.

StepStatusWhenEvidence
Demonstrated
shown at least once
yesFeb 27, 2024Klarna said its assistant handled 2.3 million chats, two-thirds of the total, in its first month, in 35+ languages; these are the company's own figures. (VKTR (Klarna release))
Reproduced
independently repeated
yesSep 3, 2025Salesforce's CEO said AI agents handled 50% of its customer service interactions as it cut about 4,000 support roles. (TechRepublic)
Accessible
available to ordinary people, or criminals
yesSep 12, 2025A Gartner survey found 11% of companies had cut staff because of AI and 22% had stopped backfilling departures, though none of the largest were predicted to drop humans entirely by 2028. (The Register)
Reliable
works most of the time for a defined task
partlyMay 9, 2025Klarna's CEO said its AI-first approach gave 'lower quality' and it was rehiring humans; Commonwealth Bank reversed AI-linked job cuts in August 2025. (Entrepreneur)
Consequences
real-world use or harm documented
yesJun 1, 2026Attackers talked Meta's AI support agent into attaching new emails to high-profile Instagram accounts; in 2024 Air Canada was held liable for its chatbot's wrong advice. (404 Media)

From the record

  • A British Columbia tribunal held Air Canada liable after its website chatbot wrongly told a grieving customer he could claim a bereavement fare afterwards. The case file → (Civilresolutionbc)
  • Klarna's CEO said its AI-led customer service had produced lower quality and the company was hiring human agents again. (Entrepreneur)
  • Commonwealth Bank of Australia reversed 45 customer service job cuts linked to an AI voice bot and apologised, calling the decision an error. (Information Age (ACS))
  • Hackers took over high-profile Instagram accounts by asking Meta's AI support agent to attach a new email address, which it did without adequate checks. The case file → (404 Media)

What it means for you

When you contact a company you will increasingly reach an AI first. It can be fast for simple questions, but it can be wrong, and the company, not the bot, is responsible for what it tells you.

What to do

  1. Ask for a humanIf the bot can't solve your problem or the stakes are high, ask for a person; Klarna and others say one should be available.
  2. Keep a recordScreenshot or save chatbot answers about refunds, fares or policies; Air Canada had to honour what its bot said.
  3. Check important answersConfirm policies on the company's official pages before acting on a chatbot's word.
  4. Lock down your accountsUse strong account recovery settings and two-factor login, since AI support agents have been tricked into handing over accounts.

Questions people ask

Did Klarna replace its customer service staff with AI?

Partly. Its AI handled two-thirds of chats in 2024, but in May 2025 its CEO said quality had suffered and it began hiring humans again.

Is a company responsible for what its chatbot says?

In Canada, yes: a tribunal held Air Canada liable in 2024 for its chatbot's wrong advice and rejected the claim that the bot was a separate entity.

How much customer service does AI handle now?

Salesforce says about half of its service interactions; Klarna says its assistant does the work of about 853 full-time staff as of late 2025.

Will companies get rid of human agents entirely?

Gartner predicts none of the Fortune 500 will fully drop human customer service by 2028, and that half of firms planning big cuts will abandon them.

Related

Sources

Spot an error or something out of date? Tell us; corrections are made and logged. Written with AI assistance from public sources; every fact checked and approved by Colin Balfe. General information, not legal, medical or financial advice.