Most Customers Still Want Humans to Solve Their Problems
Seventy-nine percent of Americans prefer speaking with a human agent over an AI system, according to SurveyMonkey data. A separate PwC survey found that 86% of consumers said human interaction was important to their brand experience.
This preference hasn't shifted despite AI's growing presence in shopping and browsing. When something breaks, customers want a person to fix it.
The tension is clear: AI adoption is accelerating across customer operations, but customer expectations remain anchored to human contact. Leading Business Process Outsourcing (BPO) companies are responding by redesigning how support systems work behind the scenes-using AI to assist human agents rather than replace them.
The "AI First, Human Backed" Model
Global outsourcing companies are building what amounts to a two-layer system. AI handles repetitive intake work. Humans handle resolution.
In this structure, AI processes ticket classification, language detection, urgency prioritization, and response drafting for common queries. It surfaces relevant knowledge articles so agents don't waste time searching mid-conversation.
This frees agents to focus on interpretation and context. Funmi Mide-Ajala, Director of Customer Support & Digital Operations at Hugo Inc., illustrates the difference: "A customer who says 'I want a refund' might actually be telling you they feel unheard. AI will process the refund, but a skilled agent can read between the lines, pick up on emotional subtext, and get to the root of the matter."
The customer experiences a fully human interaction, even though multiple AI systems work in the background.
Routing and Triage Are the Real Design Problem
AI adoption is widespread. About 88% of organizations use AI in at least one business function, according to McKinsey & Company.
But adoption doesn't equal execution. Support quality now depends on routing decisions-whether a ticket flows through AI-assisted paths or escalates directly to a human.
The intake layer works in stages. Chatbots confirm identity, pull up order details, and collect initial information before the ticket reaches an agent. Routing systems then evaluate intent, sentiment, and complexity.
Escalation is built into the design. If the system detects frustration or complexity that suggests human judgment is needed, the ticket routes to an agent immediately. This triage model allows AI to improve speed without removing the human layer.
What the Numbers Show
Hugo worked with a 107-person support team handling high ticket volume and rising costs. Instead of layering AI on top of existing workflows, the team redesigned intake, classification, routing, and response drafting from the ground up.
Results: 60% of tickets resolved without human intervention, a 42% reduction in support costs, and a 12-point increase in customer satisfaction scores.
Deloitte research confirms the pattern. Service teams see the biggest gains when AI is built into workflows rather than added as standalone automation.
Three Priorities for Implementation
For teams building or updating customer support systems:
- Define routing logic before scaling AI. Know which ticket types go to AI and which go to humans.
- Build triage as the core system. Ensure intent, urgency, and sentiment are accurately classified before a human steps in.
- Design for protected moments. Cancellations, refunds, and escalations carry outsized weight in retention. These should reach humans quickly.
For support leaders and agencies, routing logic between AI and humans is now the core design problem. That's where experience quality gets decided.
Learn more about AI for Customer Support and AI Agents & Automation to understand how these systems work in practice.
Your membership also unlocks:
