A friend of mine recently tried to change a flight booking. He called the airline and navigated through seven menu options, answered three verification questions to a voice bot, got transferred to another bot that asked the same questions, waited 40 minutes on hold, and then got disconnected. He called back. Same process. Forty more minutes. When he finally reached a human, the change took about 90 seconds.

That’s automation done badly. The airline saved money by reducing call centre staff. But they didn’t save my friend any time, and they definitely didn’t save his loyalty.

This is the fundamental tension with automation. Done well, it frees people from repetitive tasks and gives customers faster service. Done poorly, it creates friction, frustration, and the distinct feeling that a company doesn’t actually want to talk to you.

Where Automation Excels

Let’s be clear — automation isn’t the villain here. Bad implementation is. There are entire categories of work where automation is unambiguously better than the manual alternative.

Data entry. Invoice processing. Appointment reminders. Inventory updates. Password resets. Order confirmations. These are tasks where speed and consistency matter, human judgment adds nothing, and errors from manual handling are common.

A medical practice that sends automated appointment reminders reduces no-shows by up to 38%, according to research published in BMJ. A warehouse that automates inventory counts catches discrepancies in real time instead of discovering them during a quarterly audit. These aren’t examples of replacing humans. They’re examples of redirecting human effort to where it matters.

Where It Falls Apart

Automation fails when it’s applied to interactions that require judgment, empathy, or nuance. And it fails especially badly when companies use it to create distance between themselves and their customers.

The worst offenders build automated systems that are nearly impossible to escape. Chatbots that loop through the same script regardless of what you type. Phone trees designed to discourage people from reaching a human. Email auto-responders that promise a reply within “3-5 business days” — a timeline that signals your issue is low priority.

Customers aren’t stupid. They can tell when automation is there to help them and when it’s there to deflect them. The former builds trust. The latter destroys it.

Internally, the same principle applies. When companies automate performance tracking to the point where employees feel surveilled rather than supported — monitoring keystrokes, tracking screen time, flagging “idle” periods — they undermine the trust and autonomy that good work requires.

Finding the Line

The question isn’t whether to automate. It’s where. And the answer depends on understanding which parts of your business are transactional and which are relational.

Transactional interactions follow predictable patterns. Customer orders a product. System confirms the order. Warehouse picks and ships. Tracking number sent automatically. This flow doesn’t need a human at any point (assuming everything goes to plan).

Relational interactions are different. A customer with a complex complaint. A new employee struggling with onboarding. A client relationship where trust has been damaged. These require the kind of judgment, flexibility, and emotional intelligence that current automation simply can’t provide.

The smart approach is to automate the transactional layer so that humans have more time and energy for the relational one. The team at team400.ai describes this as “automation that amplifies people rather than replacing them” — and that framing captures it well.

Practical Principles

Automate the process, not the relationship. Use automation for the steps between human interactions — data collection, scheduling, routing, follow-up — but keep the interactions themselves human when they matter.

Always provide an escape hatch. Every automated system should include a clear, easy path to a real person. Not buried in a sub-menu. Not after 20 minutes of bot conversation. Prominent and immediate. Paradoxically, customers who know they can reach a human often don’t need to — the confidence alone reduces anxiety.

Test with your least tech-savvy customers. If your 70-year-old customer can navigate the automated system without frustration, you’ve designed it well. If they can’t, it’s not ready.

Monitor the right metrics. Don’t just track cost savings and throughput. Track customer satisfaction, repeat contact rates, and escalation volumes. If your automation is generating more escalations than it resolves, it’s creating work rather than eliminating it.

Involve your frontline staff. The people answering phones and responding to emails know exactly which tasks are repetitive time-wasters and which interactions genuinely need a human touch. They’re usually right.

The Competitive Advantage of Being Human

Here’s the thing that gets lost in the efficiency conversation. As more companies automate their customer interactions, the ones that maintain genuine human connection stand out. Being easy to reach, responsive, and personally attentive is becoming a differentiator precisely because so many businesses have automated those qualities away.

A 2024 PwC survey found that 82% of Australian consumers want more human interaction in their customer service experiences, not less. That number has been climbing for three consecutive years.

The companies winning this aren’t anti-technology. They’re pro-human. They use automation aggressively behind the scenes — in operations, logistics, data processing, internal workflows — and preserve human connection at every customer-facing touchpoint that matters.

That’s not just good ethics. It’s good business.