As our fearless leader Mike Murchison has proclaimed, “AI just had its Netscape moment” — which is a fun way to say that AI is about to fundamentally change the way we do business. In fact the change has already started.

Our #LLM slack channel is lighting up almost daily with announcements from enterprises adding generative AI capabilities to their product or service, and it seems there’s nothing that ChatGPT can’t do.

While the hype is real, and the potential for generative AI is fascinating, smart CX leaders like yourself know that the tech is still limited, and know to approach companies touting LLM capabilities as a magic solution with scrutiny.

I wanted to get a clear idea of how generative AI can actually be implemented right now within customer service departments, so I asked different experts within Ada.

Here is their advice.

Can: Speed up content creation

Video transcript — Interview with Jérôme Solis, Principal Product Manager, Machine Learning:

“The first thing that comes to mind is to speed up the creation of content — using AI and the LLMs as a writing assistant, as a building assist to speed up the time to value between zero and having a live bot. We know that is doable and probably also desirable because the human is part of the loop. The AI is there to just speed up the process, but at the end there's a human to review that what has been generated is actually safe, truthful, and also helpful.

And the great thing about this, too, is that the feedback that a human is giving is something that we can capture to further improve the model. So there's really a virtuous cycle between the machine doing the heavy lifting of creating content, and reserving the human for higher value activities such as [producing] original content and making sure that everything is aligned with what the brand wants to say.”

The lowest hanging fruit is using generative AI to speed up support content creation. Companies should be using generative AI like an assistant and a copilot for the builder team in charge of developing content for customer service automation. Jim Monroe, Ada's VP of Customer Experience, adds, “generative AI allows us to be heavily dynamic, which means that these large language models can pull in answers from your existing support content.”

With AI developing the first drafts and sourcing information from content you already have, you can accelerate the pace of building automated flows and getting them in front of customers to start making an impact (read: speed up time to value).

Can: Reinvent your CX organization

Video transcript — Interview with Jim Monroe, VP of Customer Experience:

“When you look at generative AI and the way that the industry is going, and the way that Ada is innovating, that takes a lot of the pressure off of the “build” part and it really begins to emphasize more on the analytical part of things.

So it's not simply about setting up and generating initial questions and responses like an FAQ. It's about actually analyzing what is causing those interactions and what is a different type of flow or interaction framework that could mitigate these types of concerns and questions that customers have.

Then from a resourcing, tooling, and skillset standpoint, you're beginning to evolve from a bot builder into a true bot manager — and then that bot manager begins to leverage other skills or individuals. For instance, there is a premium that's placed on the conversational design that's going on. There's a premium that's placed on the data analytics, because in many cases you want to integrate and pull in data sources to personalize interactions.

There's a premium on the data science aspect, which is looking at when we force rank all of those types of interactions, how do we ensure that we're getting the biggest return on our investment for deploying AI? So it really does begin to morph. And it's less about the fundamental building and more about the analytics.”

Automation has already changed the way our clients structure their CX organizations and created new career paths as well. Generative AI will continue building on that, and CX leaders will need to keep an eye on what skill sets they already have on the team and where there are gaps they need to fill.

Can: Offer built-in conversational design best practices

Video transcript — Interview with Kayla Thomson, Product Manager:

“Advances in AI are making LLMs much more accessible to a range of individuals who might not have a background in automation. And LLMs are really great when it comes to building out conversational design and interactions and content because it innately has that conversational design practice within its responses and the content that it's generating.

One of the features that we are releasing is going to make it much easier for users of our product to incorporate conversational design best practices. We're going to be injecting alarms into the content editor to output content on behalf of our customers to have those built-in conversational design best practices.”

An example of this is that bot managers can essentially type out a few bullet points and the AI can use the built in conversational design best practices to reformat them for the channel that the conversation is happening on — email, web chat, SMS, phone call, or something else.

Can: Guide your agents strategically

Video transcript — Interview with Kristal Lam, Senior Director, Product Management

“I think a lot of this can impact the way we guide builders and agents and how they work. With generative AI, we can really take out parts of the conversation or article or the ticket that someone is working it, and be able to surface the most important thing that needs their attention.

We can flag stale content to builders so they can see how they can make improvements there. You can see specific articles that might not be hitting your goal resolution rate, and from there we leverage generative AI to give you exact pointers and words and improvements you can make to make that resolution rate go up.

You’re now able to be more strategic and [the generative AI] can flag things before you even catch the error.”

Much like Jim pointed out earlier, there’s going to be a much bigger premium placed on data analytics and data science. Generative AI can help surface a lot of insights, but it’s ultimately up to the bot managers to prioritize and act on them based on what would be the most valuable.

Can’t: A resounding “set it up and trust it completely”

A curious thing happened when I asked the question “what can’t CX leaders do with generative AI.” Without fail, every single expert paused briefly, gave their answer, then used the word “yet.”

Jérôme pushed it even further and said, “it's hard to say because with the right guard rails, there's a lot of things we can do,” and I think this sets up the common point in all the answers quite eloquently.

What all the answers had in common was this: you can’t set it and forget it, yet.

If you’re incorporating a publicly available LLM into your techstack, you need to make sure you have a deep understanding of how this technology works and what kind of guardrails you’d need to set up around it to make sure you don’t end up on one of those AI generated fails Twitter accounts.

What you can't do essentially is just deploy it, set it, and forget it, without putting some mechanism or process in place to validate that the generated output is safe, accurate, and helpful.