The Human in the Loop: Making AI Work for Customer Experience
AI can now resolve more tickets faster than ever. But that doesn’t mean it knows when to stop talking, when to listen, or when trust is on the line.
That’s something I’ve seen firsthand, leading support transformations powered by LLMs, copilots, and conversational AI. The gains in speed and consistency are real. But the risks? Just as real.
This week, I’m writing about how we’ve learned to protect what matters most—judgment—while still embracing AI’s advantages.
Where LLM-Powered AI Excels:
Instant summarization & next-best actions → Copilots help agents catch up on multi-message threads instantly.
Policy recall with precision → AI retrieves exact rules, pricing details, and terms in seconds.
Low-effort questions → Routine issues like "How do I update this?" are resolved quickly and effectively.
These are real wins—when done right.
Where Judgment Still Wins:
False confidence → LLMs may hallucinate or overstate info while sounding right.
Emotional nuance → Frustrated customers need understanding, not formatting.
Policy gray zones → Ethical decisions, exceptions, and trust-building need people.
One of the toughest moments I witnessed came when AI confidently told a customer they were wrong—except they weren’t. It wasn’t a tech failure. It was a judgment gap.
That’s when we started thinking differently.
Rethinking CX Ops with Human Judgment at the Core
In my experience implementing AI and leading teams through change, I recommend:
Redefine escalation models → Use emotional tone, ambiguity, and impact as handoff signals—not just SLAs.
Re-scope agent roles → Train agents as discerners, not just responders. Prioritize empathy, negotiation, and decision-making.
Validate AI outputs systemically → Build feedback loops that teach models based on human QA of tone, clarity, and factual accuracy.
AI doesn’t replace judgment. It creates space for better judgment.
Design Prompt for CX Leaders
Where in your journey is the AI technically right, but experientially wrong? That’s where your humans need to stay in the loop.
I would love to hear how you're balancing speed, empathy, and trust in your support and service model.
Stay thoughtful, Guneet
Disclaimer: The views expressed in this newsletter are solely mine. I am not a spokesperson for my employer, nor do I represent my employer's opinion.