Ask Onix
AI set to dominate customer service-but at what cost?
Artificial intelligence is rapidly transforming customer service, with forecasts suggesting AI could handle 80% of routine inquiries by 2029, according to Gartner. Yet while executives like Tata Consultancy Services' CEO K. Krithivasan predict a "minimal need" for Asian call centers in the near future, real-world rollouts reveal a mix of promise and pitfalls.
From rule-based bots to autonomous agents
The shift from rigid, scripted chatbots to more advanced "AI agents"-systems capable of independent decision-making-is accelerating. Unlike traditional bots, which rely on predefined responses, these agents aim to mimic human problem-solving. However, early adopters face challenges balancing efficiency with reliability.
Take parcel delivery firm Evri's chatbot, Ezra. When a customer's package went missing, Ezra confirmed delivery-only to present a photo of the parcel at the wrong address, then offered no further recourse. Evri, which is investing £57 million in upgrades, insists its "intelligent chat facility" resolves most queries instantly. Yet rival DPD's less constrained AI chatbot was pulled after it swore at users and criticized the company.
High stakes, mixed results
Gartner reports that 85% of customer service leaders are testing or deploying AI chatbots, but only 20% of projects meet expectations. Emily Potosky, a Gartner analyst, warns that while AI enables "more natural conversations," risks include hallucinations, outdated information, or outright errors. "For parcel delivery, rule-based agents work well-there are only so many ways to ask about a missing package," she notes.
Cost savings, a key driver for AI adoption, aren't guaranteed. "This is very expensive technology," Potosky adds. Success hinges on robust training data-ironically making knowledge management more critical, not less. "Generative AI can't fix poorly organized knowledge; it amplifies its flaws."
Call centers as AI training grounds
Joe Inzerillo, Salesforce's chief digital officer, argues that offshore call centers-like those in India and the Philippines-are ideal AI incubators. "Years of staff training and documentation create a rich dataset for AI to learn from," he says. Salesforce's AgentForce platform, used by clients from Formula 1 to Reddit, now handles 94% of customer interactions when offered as an option.
Early missteps taught Salesforce critical lessons. Initially, its AI agent opened tickets without empathy (e.g., skipping "sorry to hear that") and refused to discuss competitors-even for legitimate queries about integrating Microsoft Teams. Adjustments followed, and Inzerillo claims customer satisfaction now exceeds human agents' scores. The firm has cut service costs by $100 million, though Inzerillo downplays job losses, stating most affected staff were "redeployed."
The human factor persists
Fiona Coleman, founder of QStory-which uses AI to optimize call-center shift flexibility for clients like eBay and NatWest-questions whether AI can ever fully replace humans. "There are moments I don't want a digital interaction-I want to talk to a person," she says. "Can AI handle a mortgage application or debt discussion with true empathy? We'll see in five years."
Regulatory pushback may further complicate the transition. Proposed U.S. legislation would require businesses to disclose AI use in call centers and transfer callers to humans upon request. Gartner predicts the EU could mandate a "right to talk to a human" by 2028 as part of consumer protections.
"Knowledge management is more important when deploying generative AI. The idea that AI can compensate for disorganized data is a myth."
Emily Potosky, Gartner analyst
Key takeaways
- Efficiency vs. reliability: AI agents promise speed but risk errors, from hallucinations to misplaced empathy.
- Cost paradox: While AI may reduce long-term expenses, initial deployment is capital-intensive.
- Regulatory headwinds: Laws in the U.S. and EU could enforce human fallback options.
- Hybrid future: Most firms envision AI handling routine tasks, with humans managing complex or sensitive issues.
