AI agents that respect customer choice-and free your team to do real work
Here's the headline: one airline's AI agent now fields about 40,000 customer queries per day across 1,300+ topics, with a 97% success rate. That's more than 13 million conversations resolved so far, and only 3% escalated to humans. Faster answers for customers. Fewer repetitive tickets for your team.
"AI.g now handles 97% of 4 million-plus customer queries. We've saved millions, but the real win is customers aren't waiting-they're served. Fast answers for them, unlocking real problem-solving time for our people," says Dr. Ramaswamy.
What customers actually want: speed and choice
About half of customers choose the AI agent first. The other half still prefers a person. Both options are there on purpose, and that's the point-flexibility reduces friction and builds trust.
Don't force automation. Offer it. When the bot is good, customers pick it. When the case is complex, they move to a human without fighting the system.
What your support team gains
Frontline reps stop answering the same five questions all day. They shift to the interesting work-edge cases, empathy-required moments, and process fixes.
"Employees are doing things that are more value-added," says Dr. Ramaswamy. "I think it helps with morale…because now they get to contribute at a higher level."
Operational impact you can measure
- Coverage: 1,300+ question types-from booking changes to refunds.
- Volume: ~40,000 queries handled daily.
- Containment: 97% automated resolution; ~3% routed to humans.
- Outcome: Faster service for customers, meaningful time back for agents, and material cost savings.
The next step: end-to-end workflows
The team is now piloting agents that run full workflows across systems-think refunds compressed from weeks to hours. This moves bots from "answer engine" to "action engine."
The takeaway for support leaders: agentic AI isn't future tense. It's working at production scale right now. As one leader put it, "A lot of the things that are pain points for organisations-the technology to help is there today. You don't need to wait for a later time."
How to apply this in your contact center this quarter
- Start with the top 20 intents by volume and effort (booking changes, refunds, status, policy questions). Write crisp resolution flows.
- Make "Talk to a person" available at every step. Route by intent, urgency, and customer value.
- Define escalation rules: confidence thresholds, timeouts, sentiment triggers, and SLA-backed handoffs.
- Instrument everything: resolution rate, containment, AHT, CSAT, recontact rate. Run A/B tests against a human-only baseline.
- Connect systems so the agent can act: CRM, ticketing, payment, booking, identity. Read + write access with audit logs.
- Train agents for complex case handling and AI supervision. Give them tooling to correct, improve, and expand intents weekly.
- Publish clear policies on data use and error handling. Close the loop with post-resolution surveys and fail-safe fallbacks.
Upskill your team
If you're building AI-assisted support, get your org fluent fast. Explore role-based learning paths and automation resources here:
Your membership also unlocks:
