AI Customer Care Grows Up: ROI, Personalization, and the Human Option

Customer Service AI That Pays: What Support Leaders Can Use Today

Klarna's $15 billion IPO didn't just raise capital. It put a spotlight on a bigger shift: AI belongs in customer service, but only when paired with people.

The company's first attempt leaned too hard on automation. Customers felt it. Klarna's pivot to a blended model-AI for speed, humans for empathy-reflects where support is headed. As CEO Sebastian Siemiatkowski put it, customers should always have the option to talk to a person.

Across financial services and retail, teams are moving from pilots to production. The leaders share one mindset: prove ROI fast, or shut it down.

Five Ways to Deploy AI in Customer Care (That Actually Work)

• Proactive Issue Resolution
  Flag and fix problems before customers complain-declines, fees, delivery delays. This requires stitching data across payments, logistics, and support. Do it right and call volume drops while satisfaction rises.
• Hyper-Personalized Support
  Use real-time data to tailor repayment options, offers, and loyalty incentives. Walmart has deployed AI-driven personalization across its app and eCommerce experiences, showing what "relevant by default" looks like. See their direction here: Walmart's AI experiences.
• Multilingual, 24/7 Coverage
  Always-on chat and voice, across languages, is table stakes. New multimodal systems handle voice, text, and images in one flow-think damaged items, address issues, or receipt uploads without channel hopping.
• Sentiment Detection and Escalation
  Let AI read tone and phrasing in real time. Route heated conversations to senior agents, surface empathy prompts, and protect brand trust. Companies using empathy-aware routing see fewer escalations and smoother saves.
• Insights Beyond the Call Center
  Complaints are product feedback in disguise. Run analytics across tickets, reviews, chats, and social to fix checkout friction, packaging, or policy gaps. Close the loop and you prevent the next wave of contacts.

Proof Points Leaders Care About

ROI is the filter. Projects that don't move hard numbers won't scale. Todd Fisher, CEO of CallTrackingMetrics, reported that users rated Webex AI Agent as equal or better than a human 72% of the time-and saw an 85% reduction in agent escalations, a 22% cut in average handle time, and a 39% lift in CSAT.

If you're evaluating options, start with a controlled pilot in one high-volume intent (billing, order status, password reset). Measure weekly. Expand only when the data earns it. For a deeper view of AI in contact centers, explore: Webex Contact Center.

Metrics That Decide Funding

• Containment rate: % of intents resolved without a human. Target: improve 5-10 points per quarter.
• Average handle time (AHT): Reduce by 15-25% via better triage, summaries, and knowledge surface.
• First contact resolution (FCR): Up and to the right-no repeat contacts within 7 days.
• Escalation rate: Cut by routing sensitive cases earlier and arming agents with context.
• CSAT/NPS: Maintain or improve as automation grows. If scores dip, pull back and tune.
• Cost per resolution: Track end-to-end, not just license cost. Include deflection and prevention wins.

Implementation Playbook for Support Teams

• Unify your data. Connect CRM, order, payment, and logistics. No unified view, no personalization.
• Automate the obvious first. Password resets, delivery ETAs, returns. Add safe fallback to a human in two clicks.
• Add intent routing + summaries. Use AI to extract reason-for-contact and generate agent-ready summaries.
• Enable multilingual coverage. Offer instant translation and native-language bots with "talk to a person" always visible.
• Deploy sentiment alerts. Escalate high-risk messages to senior agents with suggested responses and offers.
• Close the loop weekly. Feed top issues to product, ops, and policy owners. Publish fixes back to agents.
• Governance matters. Set quality checks, bias reviews, and model performance thresholds. Keep transcripts for audit.
• Upskill your team. Train agents to work with AI tools, not against them-prompting, verification, and exception handling.

Case: Rent the Runway's Personalization Push

Rent the Runway is rolling out personalized recommendations based on favorite designers, styles, and occasions. The goal: make picking effortless and reduce choice fatigue.

They're also using AI to surface insights from member reviews, summarize fit guidance, and improve search and browse. Alongside a major inventory investment-2,200 new styles and 56 new brands this year-the company reported growth in revenue, active subscribers, total subscribers, and an average subscription NPS around 77.

The team is moving fast on new channels too: affiliate emails that drive brand purchases, social campaigns on Instagram, TikTok, and Reddit, plus in-person events that over-subscribed 3x. CEO Jennifer Hyman called this chapter an "IPO 2.0"-a reset built on product and customer value, not just headlines.

What This Means for Support Leaders

• Make "human-on-demand" a promise. Automation should never trap the customer. Put the live option in plain sight.
• Prioritize use cases by business impact. Proactive billing, order logistics, and returns beat vanity features.
• Instrument everything. If it doesn't move containment, AHT, CSAT/NPS, or cost per resolution, rethink it.
• Treat customer service as product research. Insights from tickets and reviews should steer roadmap and policy.

If you're building your team's AI capability, explore practical training and certifications for support roles here: AI courses by job.

Bottom line: AI pays when it makes customers feel seen and speeds up the outcome. Lead with empathy, measure with discipline, and let the data decide where to scale.