Branching Out: The Evolution of AI to “Agentic” and the Future of Customer Conversations

Let’s cut through the buzzwords. Last summer, I talked about agentic AI, and many of you asked for a simpler rundown of the evolution of AI. It’s time to move from clunky, old-school chatbots to the next generation of customer experiences.

Think back to those old chatbots. They were an obstacle course of rigid menus and flawed Natural Language Understanding. If you didn’t use their secret language, you were trapped in an endless loop. The experience was impersonal and never felt like a real conversation.

The evolution of AI timeline took a massive leap forward in late 2022, when OpenAI’s ChatGPT arrived. Suddenly, AI could chat like a human. It could listen, understand, and respond naturally to complex questions. I think many of us felt a mix of awe and nervous excitement the first time we used ChatGPT. It changed the game.

This breakthrough didn’t just fix old chatbots; it flipped the entire customer journey. It’s a key moment that answers the question of how AI evolved from simple tools into true partners. Customers no longer have to master outdated phone trees or confusing apps. It is now our job as businesses to understand our customers, not the other way around. With AI that can read, write, and even show empathy, interactions can feel natural.

What is agentic AI? 

It’s the next generation of AI that can think for itself within safe boundaries. The word “agentic” hints at its ability to take agency. These systems make decisions, adapt to a customer’s needs, and genuinely help. No more forcing people through scripted responses. This rapid evolution of AI is the unlock provided by Large Language Models.

Let’s make this clearer with an example.

Imagine you’re at the airport and get a text that your connecting flight is significantly delayed. Now begins the dreaded maze of rebooking. With an old chatbot, you’d go to the website and click through a maddening series of menus. If you dared to type a real message, the bot would give a polite, worthless response that it doesn’t understand. You’d then spend five minutes trying to escape the bot to find a human, only to start the frustrating process all over again.

Now, let’s try that with agentic AI.

You open the web chat, and the AI agent immediately knows who you are and why you’re there. It apologizes for the delay, shows some empathy, and proactively offers solutions. It can rebook your flight, find a hotel, schedule an Uber, and issue meal vouchers. It has access to the airline’s backend systems and can find you perks like an airport lounge pass. The agent works on your behalf to make a bad day less awful. You pick one of the three flight options it found, and it confirms the change. It also sends you a lounge upgrade with a complimentary meal voucher while you wait. Within minutes, your airline app pings with the confirmation.

The evolution of AI to now.

The journey here wasn’t without a few bumps. Early in 2023, we all heard stories about AI “hallucinations” and some PR nightmares when things went badly. Those early challenges taught everyone valuable lessons. Today, the experience is much smoother and more reliable. The first ones through the door may have faced challenges, but they paved the way for the rest of us to learn without repeating those same missteps. We were there in those early days, too. Our approach was to watch and learn, which helped us sidestep some of those very public pitfalls.

I’m not here to scare you with FOMO. You aren’t being left behind. This is simply an opportunity to evolve. Your competitors are already exploring these capabilities to deliver empathetic and efficient customer interactions. You don’t have to be the first, but you can learn from those who went before. I have my fair share of battle scars, I assure you, and some hurt a lot.

For instance, we once thought we could scrape PDF user manuals to extract relevant data for an LLM. This was a completely backward way to tackle the problem. We learned the hard way about the foundational value of clean, structured data. We also saw other brands suffer PR nightmares with their AI. That made us double down on hallucination detection and guardrails from day one. It was also clear that LLMs trained on public data inherit public biases. That lesson pushed us to build post-LLM checks to ensure every response is free of bias and stays on brand.

If you’re ready to offer a smarter, more natural experience for your customers, now is a great moment to consider making the switch. We’ve moved beyond asking if these tools work. We’re now focused on how they work best. Best practices like tool calling and frameworks like Model Context Protocol (MCP) have matured. The LLMs themselves are far more capable than they were two years ago and can be trusted to execute complex tasks. Frankly, we’ve all just gotten much better at prompt engineering and leveraging AI.

AI’s evolution is an invitation for you to evolve your CX.

Many of us have heard the adage that the best time to plant a tree was 20 years ago, and the second-best time is today. That’s not true in this space. Planting an AI tree “20 years ago” meant your poor forest had to endure unforeseen droughts, floods, and forest fires. We are now in a place where the conditions are just right, and getting better every day. Models are faster, context windows are larger, reasoning has improved dramatically, and costs are coming down.

Find a partner who can break down the business benefits in everyday language. They can guide you smoothly from outdated chatbots to an AI that truly works for you and your customers.

Now that the conditions are right, we can think bigger than just planting one tree and hoping it survives. Let’s work together to build a thriving forest—an ecosystem of smarter, more human experiences that’s built to last.

Frequently Asked Questions (FAQs)