ChatGPT and other LLMs may be the end of chat bots

Some years ago, people invented chat apps that allowed you to be able to start chats online to make enquiries and get support. Initially, there were real people on the other side of the chats but over time, the issues with this became clear: sometimes the people meant to pick up the chats weren’t available or the quality of responses between different support agents was inconsistent.

So, people came up with chatbots. The bots chat with people and if they’re unable to resolve the issue, then the chat is routed to a real person to handle. But even with this, there was still a problem. Chatbots are invariably limited to their programmed pathways so if you ask questions in ways the bot isn’t programmed to process, then the system fails miserably and you don’t get the help you came for. 

Unfortunately, almost all the chats were going this way, the obviously wrong way.

This is similar to what happened to the interactive voice response (IVR), those automated phone systems where you’re told to press 1, press 2 etc. depending on what you want. This solution isn’t so intuitive, doesn’t remember details or understand context. It also doesn’t do well with typos, so, if you make a mistake with your selection, everything goes off and you have to start all over – very frustrating.

Then came ChatGPT …

In November 2022, OpenAI changed humanity’s understanding of what it is to talk to a machine or bot: ChatGPT was born and it was extremely disruptive. Almost human-like. But of course, as amazing as this new shiny technology was, it and other AI models like it, had their own issues at launch. They frequently hallucinate, which is to make up things that didn’t exist. Worse still, their responses were only limited to the knowledge they had at the time.

Some scientists even called this fabrication; utter, stupid, falsehoods.

Things were really bad then but they’re improving and now we’re seeing the rise of new chat technologies that aren’t trained in strict conversational pathways. This new generation of chats allow for conversations that are free-flowing and feel much more natural. 

Since ChatGPT’s launch,  we’ve seen it improve significantly. During its early days, you could only chat with it but now you can do so much more: browse, analyze data ,write code and even create your own custom chat that others can use. 

Initially, these advanced features were only available to paying customers but they’ve gone steps further and made it so you can access chats without signing up and making custom-built chats discoverable even by those who aren’t registered users.

For instance, in my company, Lendsqr, we experimented with the advanced features by putting some of our content together to create our own secret Lendian GPT and made it available to our staff. I have a subscription so I was able to create the custom chat, but my colleagues don’t need to have a subscription to use it. They can ask this chat natural language questions and the responses are pretty impressive. It’s not quite there yet, it still hallucinates once in a while, but I can see it’s improved significantly over time.

There’s more disruption to come from generative AI

Interestingly, as ChatGPT tries to find more ways to monetize and expand to other areas, a few changes are expected. Here are some of them:

They’re going to create embeddable custom GPTs that you can integrate on your website and on your app using an SDK. 

Secondly, we can expect to see a multimodal ChatGPT, which means beyond text chat, you’ll be able to interact with it using voice and video inputs. It would also be multilingual to eliminate any language barriers; users can speak to it and it’ll answer in the language they choose and maybe even be able to choose the voice it uses.

This all sounds great but my fear is that when all this happens, these advanced features are likely going to completely destroy all the other guys that are running chat service businesses, like tawk.to and FreshChat. Let’s be very frank, if the ChatGPTs of this world can offer natural language chat and the capability to perform actions e.g. call an API based on users prompts, then these other guys are dead in the water for sure

It’s going to create an entirely new world. 

This situation reminds me of the time when typing skills were a big deal and how many words you could type per minute mattered – one of my technical assistants can type 78/79 words per minute –  but these skills don’t mean anything anymore. 

Similarly, customer service is evolving. Embedding ChatGPT or similar technologies from companies like Anthropic or Mistral AI, on your website or app and connecting them to your backoffice to initiate actions and complete events, is the future of customer service.

And this tech is going to handle itself very well and completely redefine how service providers engage with their users. Just watch. 

What could this mean for businesses? 

There’s all this tech but it’s useless if you don’t know what to do with it. Right?

For example, with my work at Lendsqr, we have lenders signing up and a product support team to assist them where necessary. Now imagine if our custom-built AI chat becomes so good that it can handle our lenders complaints and resolve them in real-time. It wouldn’t only be able to read chats, but also read screenshots to figure out what the error is and possibly take an action that fixes the issue immediately. 

For banks, this tech means that they could deploy their chat and handle millions of customer interactions all at once. Some customers could start a chat and decide to switch to voice in their native language and have their queries translated and resolved quickly; including complex issues like chargebacks. 

All of this is going to open up a world where supporting customers will become so intuitive and fast, that if in five years time we talk about how customer service was bad, people will not be able to understand. This is what I see ahead. 

Unfortunately, I also see only very few companies dominating this space. If ChatGPT doesn’t do this now and another company builds this solution using OpenAI’s platform, the problem is that OpenAI can see that it’s working and decide to build a similar feature that destroys those other businesses. 

We’ve seen this happen a lot with the bigger companies like Apple and Amazon who’ll see a solution or product that works well and replicate at scale; killing the actual innovators.

But this transformation of customer service is inevitable. AI-driven support solutions are the future of how businesses will interact with their customers. There’s still some distance to cover between where we are now and what I envision but this is more than a prediction, it’s going to happen.