The automation genie is out of the bottle — but should you really let AI talk to your customers?
With 24/7 chatbots, instant response times, and lower costs, AI promises a revolution in customer service. But there’s a catch: when it fails — it can break trust, lose customers, or even spark lawsuits. So we asked 45 founders, technologists, legal experts, CMOs, and customer experience leaders one urgent question:
Is letting AI speak to customers a brilliant move — or a total disaster waiting to happen?
Here’s what they told us.
1. Yes, let AI talk — but only with humans nearby
Nearly every expert agreed: AI should absolutely play a role in customer service — but not alone.
“Let AI start the conversation, let people finish it,” said Richard Valente from Teleperformance.
“Use AI for routine tasks, but give customers a clear path to a human,” added Rich Atkinson from Airteam.
Whether it’s password resets, appointment reminders, or order tracking, AI shines on the basics. But the moment a query becomes sensitive, nuanced or emotional? The handoff to a human becomes critical.
2. Trust, transparency, and control are non-negotiable
“Your AI speaks on behalf of your brand — and you’re legally accountable for it,” warned Dan Pearce, General Counsel at Holding Redlich.
That means training it properly, testing it rigorously, and letting customers know when they’re talking to a machine. Multiple experts cited examples where things went wrong — including the infamous Air Canada and DPD chatbot disasters. The lesson? Use guardrails, audit logs, and escalation systems from day one.
3. AI should extend your team, not replace it
“It’s not about automation vs. humans,” said Emma Seymour, CFO at Deputy.
“It’s about letting people do what they do best — listen, solve problems, build trust.”
Many experts framed AI as a co-pilot, not a replacement. When AI handles the repetitive grunt work, staff have more bandwidth to focus on complex customer needs.
“AI can process returns. Humans can deliver empathy,” summed up Christopher Connolly at Twilio.
4. Your customers still want human connection
Even so-called “AI natives” want a human option.
“80% of young Australian AI users still want to escalate to a real person,” said Bede Hackney from Zoom.
Top complaints included generic responses, dead-end bots, and frustrating loops. The brands that thrive will be the ones that balance automation with empathy.
5. Bad AI is worse than no AI
“AI that’s inaccurate, misleading or untrained is a fast track to reputational risk,” said Sharon Nouh, CEO of ProSpend.
“Don’t rush to deploy. Train it like a new employee,” advised Victor Horlenko from Devart.
A surprising number of experts emphasized that bad AI is worse than none — especially in industries with legal or financial stakes.
6. Start with structure — then scale
“AI isn’t a gimmick. It’s infrastructure,” said Karim Rayes, Chief Product Officer at Nexxen.
From content moderation to customer segmentation to agent co-pilots, AI has moved far beyond simple scripts. But every powerful system must be observable, auditable, and explainable, said Rahul Tabeck of SolarWinds.
The bottom line?
AI can absolutely elevate your customer service — if you do it right.
Train it. Supervise it. Give customers an out. Then let your people shine where it matters most.
As Leanne Shelton of HumanEdge put it:
“Let AI handle the repetitive. Let humans handle the relationships.”
Read the full discussion here: Letting AI handle customer service: Good idea or total disaster?
More Let’s Talk episodes here.
To contribute to future episodes, subscribe to our list here.
Keep up to date with our stories on LinkedIn, Twitter, Facebook and Instagram.
