An October 2025 Gartner survey of 321 customer service leaders found that 91% feel pressure to implement AI in 2026.
But while many teams are rushing to adopt AI for customer service, almost as many are experiencing some strain with rolling it out.
Some automate the wrong interactions first, handing over complex interactions to AI agents when a human is still needed. Others remain unable to name a clear use case for AI customer agents, and lose ground on response times while competitors move faster. Still others leave their human agents on password resets and order tracking all day while they burn out on the exact work AI should be absorbing.
All 3 teams are making the same mistake: treating AI adoption as a binary choice between completely replacing human agents or failing to scale their support function with AI.
But AI can shift human capacity toward the interactions that build loyalty and increase retention. The key is deciding which interactions AI should handle and which ones your human agents should.
AI vs. human agents: assigning the right ticket to the right agent
AI is good at pattern matching, instant retrieval, and routine inquiries. AI breaks the moment a situation needs judgment, empathy, or cross-functional thinking.
Here’s a great place to start when you’re implementing your AI customer service strategy:
Capability | AI performance | Human performance |
|---|---|---|
Order tracking, frequently asked questions (FAQs), simple returns and exchanges | Provides instant answers that are available 24/7 | Slower, and unnecessary for these repetitive tasks |
Emotional de-escalation | Detects negative sentiment, but can't empathize and de-escalate | Understands emotional context, adapts their tone, and works to rebuild trust with the customer |
Complex multi-step resolution | Handles scripted paths, but breaks down on edge cases | Navigates ambiguity and exercises judgment |
Brand relationship repair | Generates templated apologies | Understands what went wrong for a customer and can adapt messaging to specific situations a customer |
Cross-functional problem solving | Limited to its training data and integrations | Pulls in context from multiple teams and systems |
According to Klaviyo’s 2026 AI consumer trends research, more than 1 in 5 people say that they feel most uncomfortable with inaccurate AI and AI that feels “too personal.” At the same time, 41% of consumers say customer service chats that “don’t feel human” are the top brand experience that feels “too automated,” or not personal enough.
Keeping humans in the loop for service interactions that require the kind of empathy only humans can express will keep your customers happy, especially when your human agents can devote more time to those inquiries because AI is handling routine volume.
Downsize your service team in 2026, and you'll likely be rehiring by 2027
Aforementioned October 2025 Gartner research found that only 20% of service leaders have actually reduced agent headcount because of AI. According to the survey, “The majority report that headcount remains steady, even as they support more customers.”
Meanwhile, according to Gartner's February 2026 research, 50% of companies that attributed headcount reduction to AI will rehire by 2027 for similar functions under different job titles.
Gartner's March 2026 research predicts that more than 50% of customer service organizations will double their technology spend by 2028 without an equivalent cut to talent.
Meanwhile, roles are expanding. The same Gartner survey found that nearly 80% of organizations plan to move agents into new positions, and 84% are adding new skills to frontline roles. Gartner's October 2025 survey found that 58% of service leaders plan to upskill agents into knowledge management roles.
Where support teams should start with AI
The brands getting real results don't try to automate everything at once. They start with the one use case that's clearly overloading their team, prove it works, then move on to the next.
For most teams, that means tackling 3 priorities:
- Routine questions such as “where is my order?” (WISMO) tickets
- Simple product questions at check-out to prevent abandoned carts
- Easy returns and exchanges
Deflect routine volume from humans to AI
Happy Wax, a home fragrance brand, uses an AI customer agent to manage routine support conversations. Within 90 days, their AI agent handled more than 50% of support conversations without any service team involvement, according to Rachel Fagan, vice-president of marketing at Happy Wax. “Customers get instant answers, and our team gains bandwidth for high-touch moments,” she says.
Give the AI agent your customer data, not just your FAQs
A support agent that only knows your help center can tell someone about your return policy. But an AI agent that’s drawing from your customer data can do far more, because it knows who's asking.
Take two common service conversations:
- A customer asks for a discount. An agent with customer data sees the open cart and the customer's loyalty tier, then applies the right discount code to close the sale.
- A returning customer asks about a product. The agent pulls their browsing behavior and purchase history to recommend the right items in real time, turning a routine question into a cross-sell.
Neither of those is possible when the agent is walled off from your customer data. This is why it’s so important that your AI customer agent is embedded within the same CRM that centralizes your data.
Run the same agent across every channel
Your customers don't stay on one channel, and your AI agent shouldn't either. The goal is to configure the agent once and have it deliver the same personalized service over web chat, email, SMS, and WhatsApp.
Two more service conversations show why that matters:
- A customer starts a return on web chat, then picks it back up over email that evening. Because the agent is reading the same profile, the second message continues the first, no order number re-entered, no story retold.
- A shopper asks about sizing on WhatsApp, buys, and later texts to check on shipping. The agent already knows the order, so the SMS conversation starts where the WhatsApp one left off.
Neither customer repeats themselves, because one agent is drawing on one customer profile no matter where the conversation moves. Run a separate agent per channel and the history doesn't travel with the customer, so they have to start over every time they switch.
3 steps to AI customer service implementation
The better starting point is your own data, specifically how much of your volume is routine versus complex.
Step 1: Audit your current volume
Break your support tickets into categories:
Category | Examples | AI suitability |
|---|---|---|
Routine, high-volume | Order status, password resets, return policy questions, and shipping timelines | High: AI resolves autonomously |
Moderate complexity | Product recommendations, sizing guidance, subscription changes | Medium: AI assists, a human may resolve |
High complexity / emotional | Billing disputes, failed deliveries, and complaints about service quality | Low: human-led, AI can provide some context to human agents |
Step 2: Determine your escalation criteria
Before you launch your AI agent, you’ll need to decide when AI interactions need to be escalated to a human.
- Define the escalation triggers. Name the conditions that send a conversation to a human, like high-risk or sensitive situations, complex multi-step issues, or anything outside the agent's defined scope.
- Create the escalation transparency. Customers should know when they're interacting with AI and when a human has stepped in.
Step 3: Invest in the skills AI can't replicate
As routine work moves to AI, the skills that matter most for your human team will likely change. Prioritize development in complex problem resolution, empathy and de-escalation, brand voice and values, and knowledge management.
Keep your AI and human agents on the same page with Klaviyo
With Klaviyo Service, turn every conversation into revenue with an AI agent that resolves questions automatically, personalized self-service, and live support that knows the full customer story—all on one platform.
- Customer Agent: An AI agent that resolves routine questions, recommends products, and drives purchases across web chat, SMS, email, and more, around the clock. When a question needs a human, the full conversation transfers to your team with complete context.
- Customer Hub: A branded, self-service destination where customers can manage orders, make returns, and discover products.
- Helpdesk: A shared inbox where your AI and human agents support customers with full context, including order history, loyalty status, and past interactions.
Ready to scale customer service with AI? Get started with Klaviyo Service.
Related content
