Artificial Intelligence

AI Trends in Customer Experience for 2023 and Beyond

Alan Finlay is head of product for OvationCXM and works with his team to infuse the latest use cases of AI into our CX management platform, CXMEngine, every day. We sat down with him for a Q&A on artificial intelligence and customer experience. In this FAQ, he shares his expertise on:

• What is generative AI and how it compares to other types of AI?

• Use cases for generative AI and customer experience 

• AI trends and predictions for all industries, but especially customer experience

• How AI will affect the future of work and the human/machine balance

• How AI is being integrated into OvationCXM’s customer experience management (CXM) platform

What is generative AI compared to other forms of AI?

Alan Finlay: Generative AI is a subset of artificial intelligence overall. It is a form of AI that specifically focuses on generating new content like text, images, audio or video. It takes a data set and turns it into something brand new.

Machine learning, in contrast, is focused on learning and recognizing patterns in data and making a prediction and classification about what comes next. It looks at all of the data, finds a pattern that matches and predicts what may come next. An example is Google search. You type in a few words, and it displays a list of options. That's machine learning trying to predict what you might type next.

Generative AI, on the other hand, is powerful because it creates brand-new, unique and creative outputs. It can answer questions, summarize large data sets, create images from descriptions and produce music based on a description of a feeling. The possibilities are endless, and when you see it in action, it does feel like magic. 

Can you share an AI customer experience example using conversational AI vs. generative AI?

Alan Finlay: Let’s compare the different types of AI by looking at an example of AI customer service chatbots or virtual assistants.

A customer comes to your website and asks a question. You understand the intent of the question, what the customer is trying to learn and then map that to an answer in your system. When somebody asks A, we usually answer B. Natural language processing creates very good question-and-answer responses.

This is great when you want to control the narrative and provide an exact message that leaves nothing to chance. The downside? It takes time and effort in the conversational design to create the messages, define the intent, train the data and make sure those experiences are truly seamless.

Compare that to generative AI, in which you feed the AI your content, which could be a knowledge base, database or public website documents. The AI generates a new and unique answer for that specific customer. 

The downside is you don't have full control over the content generated. If you ask the same question twice, you can receive two different responses. That can be risky, depending on the type of questions you're answering. The upside is you don't have to write and define every possible answer and outcome. Instead, you can focus on creating the best foundational content for the model to use, and let it answer the questions.

Like most things in life, I think the best answer is somewhere in the middle. A hybrid of these solutions. That's how OvationCXM leverages AI. We use different tools to solve different problems. 

In a hybrid solution, a company might create a structure where a subset of questions receives exact answers. It has full control so the answer flows to its exact specifications.  But maybe the team doesn't have time to write 10,000 answers to 10,000 possible customer questions. In that case, they could use generative AI to answer long-tail questions and give an excellent experience. It adds a little bit of risk, but they’ve mitigated that risk by ensuring the high-impact questions receive defined answers. 

What are the AI trends you see for the remainder of 2023 and into 2024?

Alan Finlay: That’s a tough question to answer given the speed at which everything's changing in the landscape of artificial intelligence (AI). Every week, if not every day, there's some new major breakthrough in the space or innovation coming to market that's changing what's possible, so I can talk about 2023, but in 2024, I have no idea! 

I think we will see significant advancements in generative AI and large language models (LLMs). That's one of the biggest changes that happened over the course of this past year, in the land of AI. I think the models will become more efficient. They will become higher performance. We'll see higher-quality outputs, and I think we'll be able to fine-tune them more easily and more accurately with the models available.

In addition to the models getting better, I think we're going to see new applications of the models that will uncover new uses for what already exists today. If we didn't change the underlying AI technology for some of these large language models for the next 12 months, we might see new innovations and what they're capable of. There's a lot of experimentation happening. 

It's not just the technology of the underlying models that’s going to improve. We’ll improve how we use them. What are the use cases? What's the output of these models? 

I think we'll see continued natural language processing advancement. Generative AI models will become more natural and more context-aware. They'll be able to have more inputs about everything surrounding them to deliver more personalized answers so it’s going to feel like magic.  The language generated will start to feel better than what you can personally write in a lot of cases, and it's going to contain all of the context… more than a human can actually put into their brain all at once. 

Beyond language, there's going to be advancement in incorporated actions. As they continue to evolve and new applications become available, we are going to move to questions like: 

  • How can you chain these AI models together? 
  • How can the AI models generate and complete tasks? 
  • How can they interact with other third-party software systems? 

We'll start to see AI more deeply integrated with the broader technology ecosystem,  not just a siloed feature that people might leverage. Instead, they will become an integrated piece of how systems start to interact with each other.

Why should companies invest in AI for customer experience?

Alan Finlay: AI is going to change not just the customer experience industry but every industry. If you look back 10 to 20 years at what software did in every industry, AI is going to do the same and magnitudes more.

CX will be significantly impacted by AI in a good way. It's going to make customer experiences across the board significantly better, for both the teams delivering the experiences and for the customers receiving those experiences.

The teams that deliver the experience will be equipped with better data, context and insights and faster answers at their fingertips. They're going to be able to focus on the things that humans do best rather than spending time on repetitive tasks. We'll see some of these repetitive tasks go away, like summarizing long email chains or searching through databases to find information for a customer. That will be instantly available at their fingertips so they will have an easier time delivering those experiences.

On the customer side, you will receive answers and everything you need much faster, and I think personalized experiences at scale will be much easier to achieve. It's very hard today. The context, history and customer sentiment AI models can have will be a critical part of delivering better CX experiences where everybody feels like it was personalized for them. And it will be using these new tools.

Remember the Staples easy button? Click a button - that was easy! That's what it's going to be like for the CX industry. Everything is going to get easier and better when AI tools are part of CX software.

What is OvationCXM’s position on using generative AI?

Alan Finlay: We have leveraged AI in our software for many years, from conversational AI and chatbots to AI-powered agent-assist knowledge, suggestions and knowledge delivery. We have a robust set of AI tools throughout the platform today, but the evolution of generative AI in the past year (and especially the past six months!) has changed what's possible.

We try to 1) only leverage AI when it makes sense and 2) we build in flexibility. 

When we add AI, we only do it when AI is actually the best way to solve a specific customer problem. AI is sometimes the best way, but oftentimes it is not. We try to be thoughtful about where we include it, and if we don't need to use it, we don't. AI is more complex, and it adds more variables. 

At OvationCXM, we also take a flexible approach, using different types of models from different organizations so we don't get locked into any specific one. I think this is critical, especially with how fast the market is changing. If you build on one specific instance or one specific provider and the market shifts in a new direction, you can get stuck. 

With the large language models in our platform today, you can move between Open AI's GPT or between Bard and Llama, depending on what provides the best performance and quality for that specific use case. 

What is OvationCXM’s strategy for infusing generative AI into its platform?

Alan Finlay: Our AI strategy is to take a somewhat measured approach. We work with financial institutions and enterprises, so being thoughtful about rolling out AI is really important to us - and to them. Although generative AI is powerful, it is not without risk. And one of the biggest risks in the CX world is what are called hallucinations.

These AI models generate brand new content on each use, so every question gets a totally brand-new answer. But there's no guarantee the answer or content AI is generating will be factually correct, EVEN IF it uses a factually correct, valid data set to generate the content. Probabilities are high it will be, but that’s not a guarantee.

So our platform allows organizations to leverage generative AI immediately, but to start with low-risk applications and slowly move towards higher-risk applications as they gain trust and fine-tune the generative AI models. 

We start with summarization, which is the lowest risk because it's not customer-facing; it's internal. An example is asking the model to generate a summary of all of the emails in a thread or the interactions with this customer over the last six months. 

Then we move to suggestions. Our CXM platform brings answers to your internal users using your content. An example might be: “I see you're chatting about this. Check out this article, and look at this answer because it might be helpful.”  Or you can type in a prompt/query, and the AI delivers the specific answer directly back to that front-line agent. There is still human oversight of the AI because it’s not delivering anything directly to a customer. Teams can review, edit and control what goes out while saving a significant amount of time. 

Next, you can improve the suggestion quality by reviewing suggested answers in the training consoles and editing them. The feedback loops make the models better, and the more fine-tuned they are, the better answers they help teams deliver.

Once response quality improves with internal users, it can be extended to answer customers directly. For instance, if a customer email comes in, AI can generate the response. Initially, somebody reads, reviews and sends.  Eventually, when they're clicking send over and over, it can begin to automatically send the email. 

Finally, you leverage insights. There is a lot of data and messages flowing, so naturally you want insights into what's going on. With AI, you can ask your data questions in natural language and get answers and charts back without heavy data analysis. For instance, “What are the most escalated topics? What are customers most concerned about?  

What do you predict for AI and the future of work? 

Alan Finlay:  Generally, we view AI as a tool to empower humans so they don’t have to think about the mundane but can, instead, focus on doing what humans do best: being thoughtful, critical thinkers; having empathy and ensuring customers are taken care of.

Examples are delivering answers to CX teams so they don’t have to search for them. Or suggesting article topics to write about based on customer questions vs. poring over Excel spreadsheets to find trends in the data.

It will empower users delivering cx. That being said, there will be areas where full automation will take shape and grow as time passes. But I don’t believe the role of the human will go away. Rather, it's going to evolve. It’s going to move to a different focus and new type of work leveraging a new suite of tools. 

For example, someone might spend their day reviewing, editing and approving content in a training console rather than writing it from scratch. That content will still need oversight and the human brain to ensure everything is factually correct when it's going out the door.

In the customer experience industry, leaders will become intimately familiar with AI, and I think they'll view it as a critical part of their toolkit to deliver high-quality and personalized experiences at scale. It will feel as common as email or Slack, and in 10 years, we'll look back and wonder how we managed CX without it. 

I think there will come a time when software that is not infused with AI will feel antiquated and hard to use. If you've tried typing on a mobile keyboard lately without auto-correct, I think it will feel like that. You are typing away furiously and look down, and nothing makes sense! You'll realize how dependent you've become on it, and you don't even know it's there! It's invisible to the human eye, helping you along the way. 

What is the risk for companies that don’t use AI for customer experience management?

There will be plenty of companies that resist bringing on AI. 

The risk of sitting on the sidelines won’t be immediately apparent. But as time goes on, customers will become accustomed to faster, better and more personalized experiences from AI-powered companies. The companies that do not adopt AI will be leapfrogged by the ones that do. I believe it will start to feel difficult to engage with companies that don’t leverage AI. 

Companies that adopt AI will be able to take the time, people and brain power they save and focus it on growth and strengthening their value proposition. I think they'll have higher NPS scores and better customer retention, and they'll ultimately take over market share. 

AI + CX is going to be a critical linchpin for growing and maintaining companies and being leaders in an industry. It will become very hard to retain customers if you don't have AI.

What are the future plans for Generative AI in OvationCXM's platform?

Alan Finlay: I have a pretty good idea of what we're going to be building over the next three to six months, but it's almost impossible to predict much further out than that. I can say with confidence that generative AI will continue to have a strategic place on our roadmap into the foreseeable future. 

We will keep our finger on the pulse of the market, the new technology and our customer needs, and quickly adapt so we can deliver high-impact products that enable the best possible customer experiences for end users.

Watch the full Q&A with Alan on artificial intelligence and customer experience.

Click to download our AI eBook!