The human touch: A CIO's wake-up call on customer service AI
AI is everywhere in customer service. It trims costs and speeds up simple tasks. But if you're leading a support team, you've felt the downside: frustrated customers, rising churn risk, and a brand that feels colder by the quarter.
Great service still runs on simple truths: respect time, show empathy, communicate clearly, and make the path to resolution obvious. Cut too many humans out of the loop, and those basics collapse.
What happens when you over-automate
Over the last few years, big brands have reduced support headcount while leaning hard on bots and IVR. The results aren't hard to predict.
- Atlassian: 350 customer service and support roles cut in 2025, citing AI replacement.
- Klarna: 700 roles cut from 2022-2024; later admitted overreach and began rehiring to fix quality issues.
- Salesforce: Customer support team reduced from 9,000 to 5,000 in 2025.
- Sky UK: 2,000 customer service jobs cut and three sites closed in 2025.
I saw the ripple effects firsthand. After a week of unstable utility service and a string of automated updates, I tried to call. The system blocked me from reaching a human. Live chat? Same wall. I went to a field office and finally spoke to a rep who shrugged and hinted the call center was gutted. No empathy. No ownership. Just a system that made me feel invisible.
What customers actually want
Customers prefer people, not bots-unless the bot gets the job done fast. In a 2025 Verint report, 56% preferred human agents, though 44% were open to automation if it solved the issue. A 2023 Gartner survey found 64% would rather companies didn't use AI for service, and 53% would consider switching providers if they did.
The top fear: you can't reach a real person. Only 13% of customer journeys start and finish in pure self-service. For the rest, forcing a bot-only path is a dead end.
The profit trap: short-term savings, long-term loss
McKinsey estimates AI in call centers can cut costs by up to 30%. That's real money. But if the trade-off is churn, you lose more than you save.
It costs five to 25 times more to acquire a new customer than to keep one. Companies have a 60%-70% chance of selling to an existing customer versus 5%-20% with a new one. In 2024, Gartner reported 73% of chief sales officers were prioritizing growth from existing customers, and 57% ranked retention and account growth among their top three goals.
Translation: loyalty pays. Cold automation erodes it.
A practical playbook for support leaders
Use AI to assist your team and speed up resolution. Don't let it become a wall between customers and help. Here's a tight model you can implement:
- Set non-negotiables: Make "talk to a human" available within two steps in every channel. Never trap customers in loops.
- Define human-first scenarios: Outages, billing errors, cancellations, vulnerable customers, escalations. Route to people immediately.
- Use AI for the right jobs: Triage, intent detection, data lookup, summaries, after-call notes, suggested replies, quality review. Keep judgment and empathy with humans.
- Preserve context: Pass full history from bot to agent-transcript, customer profile, previous attempts. No repeating details.
- Measure what matters: Track time-to-human, first contact resolution (FCR), customer effort score (CES), repeat contacts, and churn signals. Review bot containment only alongside NPS/CSAT and retention.
- Cap automation: Limit bot interactions to two turns for complex intents. If confidence drops or sentiment turns negative, escalate.
- Keep a senior "SWAT" queue: Fast-lane tough cases to your best agents. Give them authority to fix problems on the spot.
- Be transparent: Clearly label bots, set expectations, and confirm handoffs. No pretending to be human.
- Invest in people: Train for empathy, active listening, and resolution frameworks. Reward outcomes, not handle-time alone.
- Protect channels: Maintain phone, chat, and email. Don't hide your number. Customers notice-and they remember.
Policy and guardrails you can publish today
- Promise: "You can reach a human in under two minutes during business hours."
- Bot limit: "After two failed attempts, we escalate to a person automatically."
- Clarity: "We'll tell you when you're chatting with a bot and when a human joins."
- Resolution focus: Optimize for FCR and customer effort, then cost. The order matters.
The bottom line
AI should supplement your team, not replace it. Customers want speed and empathy, in that order. Give them both by using automation to reduce friction-and humans to own outcomes.
Most important: always offer a clean exit from the AI loop to a human. That's how you protect loyalty, brand trust, and revenue.
Level up your support team's AI skills
If you're training agents and leads to work alongside AI-prompting, summarization, QA, and escalation rules-these resources can help:
Your membership also unlocks:
