How AI Agents Are Changing Customer Support

Most "AI-powered" support tools do one thing: they show your customers a help article and hope for the best. If the article doesn't solve the problem, the customer ends up waiting in the same queue they would have been in anyway — just more frustrated.

GudDesk takes a different approach. Our AI agents are designed to actually resolve conversations, not just deflect them.

The Difference Between Deflection and Resolution

Deflection is when an AI says "Here's an article that might help" and closes the chat. If the article doesn't answer the question, the customer has to start over.

Resolution is when an AI understands the context of the conversation, takes the right action, and confirms the issue is solved — or seamlessly hands off to a human when it can't.

GudDesk's agents aim for resolution.

Meet the Built-In Agents

GudDesk ships with three agents out of the box:

GudBot — The First Responder

GudBot handles the initial customer interaction. It reads the message, searches your knowledge base, and responds with relevant information. But here's the key difference: GudBot tracks whether the customer's issue was actually resolved. If the customer says "that didn't work," GudBot either tries a different approach or escalates to a human.

TriageBot — The Router

Not every conversation needs AI. Some need a specific team member. TriageBot analyzes incoming conversations and routes them to the right person or team based on content, customer history, and your configured rules. Billing question from an enterprise customer? Goes straight to your account manager. Bug report? Lands in the engineering queue.

OnboardBot — The Guide

OnboardBot activates for new users. It proactively checks in, offers helpful resources based on where the user is in their onboarding journey, and flags users who might be struggling to your customer success team.

Building Your Own Agents

The built-in agents cover common patterns, but every business is different. GudDesk's Agent API lets you build custom agents that integrate with your specific tools and workflows.

Here's what a basic custom agent looks like:

import { GudDeskAgent } from "@guddesk/agents";
 
const refundAgent = new GudDeskAgent({
  name: "RefundBot",
  trigger: "intent:refund_request",
  async handler(conversation) {
    const order = await lookupOrder(conversation.customer.email);
 
    if (order && order.isRefundable) {
      await processRefund(order.id);
      return conversation.resolve(
        `Your refund for order #${order.id} has been processed. You'll see it in 3-5 business days.`
      );
    }
 
    return conversation.escalate("refund-team");
  },
});

This agent listens for refund requests, checks if the order is eligible, processes the refund automatically, and only escalates to a human when it can't handle the case.

Why This Matters

The average customer support interaction costs businesses $5-12 when handled by a human. AI agents that actually resolve issues — not just deflect them — can reduce that cost while improving customer satisfaction.

But the real win isn't cost savings. It's speed. When a customer gets their password reset in 8 seconds instead of waiting 4 hours for a reply, that's the kind of experience that builds loyalty.

Getting Started

AI agents are included in every GudDesk installation — free tier included. No API keys required for the built-in agents (they use your knowledge base as their source of truth). For custom agents that call external APIs, you'll just need to bring your own API key.

Check out our agents documentation to get started.