AI answers the easy tickets now. Support just became an expert job.

Most support queues were never mostly hard problems. They were mostly the same twenty questions on repeat: where's my order, reset my password, how do I get a refund, what does this error mean. That repetition is exactly what AI eats first, and the numbers from the companies that moved early are blunt. Klarna reported its AI assistant handled 2.3 million conversations in its first month, the workload of 700 full-time agents, and cut average resolution time from 11 minutes to under 2.

The economics follow the volume. Gartner projects conversational AI will cut contact-center agent labor costs by $80 billion by 2026. When the repetitive tier costs that much less to run, no company keeps humans typing the same refund macro a hundred times a day. The tier-1 ticket, as a human job, is ending.

The queue that's left is harder, not smaller in importance, which is the part the headlines miss. Strip out the easy tickets and what remains is the customer whose payment failed twice during a move, the enterprise account threatening to churn, the edge case three teams can't reproduce. That work just became the whole job.

Klarna is the useful case precisely because it went furthest. After betting publicly on AI-first support, the company walked it partway back in 2025 and started hiring human agents again, saying in its own statements that customers wanted the option of a person, especially on hard cases. The same company that proved AI clears tier-1 volume also proved the human tier doesn't automate. Both findings are real, and the second one is the career signal.

What appreciates now is everything tier-1 never required: de-escalating someone who is genuinely upset, making a goodwill-credit call when policy doesn't quite fit, spotting that fourteen tickets this week describe the same broken checkout flow and getting that to the product team before next week's release. Support has always been the company's richest product-insight channel; it finally has the time to act like it.

The day-to-day becomes brief, edit, verify: brief the AI with account context so drafts start right, edit responses so they sound like a person who read the thread, and verify anything involving money, policy, or a promise before it ships. It's the same shift we've tracked in sales, where volume went free and judgment became the job. And as in every field, the entry-level rung is the one that's shrinking, so the path in is arriving AI-fluent and learning escalation judgment fast.