Ada Support

Why do customer service chatbots still suck?

Mike Gozzo
Chief Product & Technology Officer

It’s hard to believe that OpenAI released ChatGPT just over a year ago. For many of us, it’s become an essential everyday tool, so much so that it might have you wondering: How did we ever live without it?

ChatGPT not only made AI more accessible to the masses, it proved what we, as professionals working in the space of AI for years now, already knew. Truly intelligent AI can create human-like conversations. For customer service specifically, it has the power to automatically resolve customer service inquiries and deliver extraordinary customer experiences.

Despite this, chatbots fail to deliver on this potential. Even with landmark AI developments like ChatGPT in the zeitgeist, chatbots are falling short. Let’s face it, they still suck. The reason for this is simpler than you think: a chatbot (even one powered by AI) is still a chatbot.

You need an AI agent for customer service.

Unlike chatbots, AI agents are designed to actually reason through problems, learn from interactions, and make decisions. Making the move from chatbot to AI agent for your customer service can be intimidating, but we’re here to break it down for you. It requires not just a technological leap, but a conceptual shift in how we envision AI’s role in shaping customer service. So let’s get into it.

The problem with customer service chatbots

First of all, they’re stuck in the past. They’re designed for obsolete measures of success and a bygone era of customer service.

ChatGPT made conversational AI interfaces more human-like, but its intelligence operates within the confines of pattern recognition and response generation based on a vast dataset. If it isn’t applied correctly, it behaves like a stochastic parrot, guessing at what a response should be without being grounded in a thorough understanding of the problem at hand.

Chatbots excel in sourcing information and workflows from their training data and presenting it in a user-friendly manner. But their capacity for reasoning is non-existent. They rely fully on human instruction.

We need to start thinking beyond simply evolving the existing technology, and go back to redefining what success means. Only then can we fully understand what we need to build to achieve it.

From ChatGPT to AI agents: Understanding the leap

To reiterate: AI agents are designed to intelligently reason through problems, learn from interactions, and make decisions. They can act autonomously, apply learned knowledge to accomplish tasks, and solve problems in real-time—making your customer service go from ordinary to extraordinary.

A chatbot is a passive tool waiting for our input. AI agents are proactive team members in your customer service org, capable of understanding our needs and helping us make the most advantageous decisions.

The benefits of using an AI agent for customer service goes beyond its effect on the customer experience — it redefines career paths for customer service teams, allowing them to reallocate human capital to more impactful roles. They can focus on areas like customer relationship management and preventing churn, and use insights generated by the AI agent to inform business decisions on pricing, product, and more.

The generative AI toolkit for customer service leaders

Practical guides to evolve your team, strategy, and tech stack for an AI-first world.

Get the toolkit

Let’s compare chatbots to AI agents in a language everyone can understand — food.

The recipe book: scripted chatbots

Imagine a cook who strictly follows recipes from a book, to a tee. Each dish they make is based on predefined instructions. If an ingredient is missing or a diner has a special request, the cook is at a loss.

This is how traditional chatbots operate. They’re based on scripted responses, unable to deviate from their programming to address complex or unanticipated customer needs effectively.

The master chef: AI agent

Now, picture a master chef. This chef understands the principles of cooking, can adapt to available ingredients, and personalize dishes to meet diners’ unique tastes and dietary requirements. They learn from feedback, improving their culinary skills over time.

This is the essence of an autonomous AI agent. It understands context, adapts to new information, personalizes interactions, and learns from each one to provide a superior customer experience.

For customer service, moving from traditional chatbots to AI agents means going from merely following recipes to embodying the skill, adaptability, and creativity of a master chef.

Hungry for more? Let’s address some burning questions.

Is an AI agent just ChatGPT in different clothes?

To make this comparison, we have to start by defining success. When chatbots were conceived, the definition of success was to deflect tickets away from customer service agents. Chatbots do this, and they do it well. But it’s part of the reason why they suck, and why customers still prefer to speak to human agents.

Instead, let’s evaluate success for human agents.

  • Did they follow established procedures?
  • Did they capture sufficient context to diagnose the issue?
  • Were they empathetic and aware of the customer’s emotional state?
  • Did they identify and act on relevant upsell opportunities?

Human agents are evaluated on the quality of their conversations. Success isn’t defined by deflection or containment but by true resolution.

Your AI agent for customer service should be held to the same standards — evaluate an AI agent like you would a human. That’s how you build and leverage AI to maximize business results. It’s how you get your money’s worth. It’s also the principle that guides our design thinking at Ada. More on that later.

How can you tell the difference between a chatbot and an AI agent?

If you’re striking up a conversation with what appears to be a chatbot, here are the signs you’re actually talking to a customer service AI agent:

  • It can handle multiple intents in the same question. To test this, ask about two different topics in the same question. An AI agent should be able to detect that and answer both.
  • It provides a consistent resolution to the same issue, even if it’s expressed in different ways. To test this, ask the same question in two different ways.
  • It understands the context of the conversation and correctly answers follow up questions. To test this, ask a question, wait for the response, and then ask follow up questions. An AI agent should be able to respond to them in a context of the whole conversation.
  • It addresses general questions that are not related to any specific topic in a similar way to a human agent. To test this, ask simple questions that are related to the product or service but would not have a specific answer prepared for them. An AI agent should respond conversationally.
  • It can answer questions, perform actions, and provide the same conversational experience regardless of which product line they’re being asked about, which channel they’re responding on, or what time of day it is. To test this, ask about different product lines or try the chat outside of business hours.
  • It returns answers that are relevant to the customer's question. To test this, ask an unusual question. An AI agent should still be able to give a relevant response.
  • It understands the meaning behind the customer’s query. To test this, ask a question that includes a potentially tricky keyword, such as “can I get a product for free?” An AI agent’s answer should still be relevant.
  • It provides specific answers, not just returns relevant links. To test this, ask a question that has a lot of help doc articles. An AI agent should provide an actual answer instead of simply returning a link to that article.

If you’re evaluating vendors to level up your customer service, a deal breaker should be the ease of management. AI agents for customer service should be autonomous — meaning you shouldn’t be spending any time on creating workflows or manually scripting answers. It should be able to understand and resolve issues without explicit instructions.

Another green flag to look for is reasoning.

Rather than trying to match a scripted response to a unique problem, at Ada we build AI that reasons. By doing this, we’re able to resolve at twice the rate of the previous generation.

Ada’s AI Agent draws on existing knowledge sources and data from third-party tools to reason through each situation. It considers past actions and the context of the conversation to determine how to best resolve an inquiry and deliver a truly personalized resolution.

It’s time to meet your AI agent for customer service

AI agents are not just the next step in the evolution of chatbots, and they’re definitely not ChatGPT in a chatbot costume. They’re a transformative leap in software technology itself.

This new generative of AI is poised to redefine our interactions with digital systems across various fields. The traditional boundaries of what software can do are being redrawn, paving the way for a new class of applications across every field.

This is how we need to think about AI agents for customer service. Not just in terms of how they can do their current job better, but what else, what more, they can do.

How to interview an AI agent

Looking to better understand the difference between chatbots and AI Agents? Download this guide to become an expert on the topic, and discover the success criteria you should be testing to get the most ROI.

Get the guide