AI ChatBots are Running RAGged
Customers like ChatGPT Because it’s a Product
Today, companies touting RAG improvements are a dime a dozen. But how many use cases out there even require advanced RAG when out-of-the-box solutions are already so capable? What can they promise that OpenAI or Anthropic can’t do with zero lines of code?
OpenAI demoed the worlds first multi-modal audio in, audio out capable model. Which Anthropic responded in kind with Claude 3.5t Sonnet delivering perhaps the most impressive text-to-artifact output we’ve seen so far. But Anthropic is nothing more than a tech-geeks secret sauce. OpenAI is still dominating the industry with a 69.9% of market share, with no other paid tool owning more than 5.2%. what does that say about the industry?
ChatGPT is king for a reason. it’s generic, it’s open, and it doesn’t require as much knowledge on what you can and cannot do. Just start using it, and it’ll guide you along the way.
It’s also a product. Consumers go to ChatGPT for a specific reason: to use the ChatGPT product. Consumers don’t come to your site to use your companies chat agent. They want your PRODUCT. They want it to be easier to use, enjoy, or take advantage of your product. They don’t want a new process to learn.
Creative Use Cases Aren’t your Friend (Yet)
Let’s take a company like Autozone as an example. They may be thinking that it would be a great idea to build their own version of ChatGPT. Customers can tell this agent what their car problem is and it will spit out the right products for the job. Sounds like a good idea right?
But you know what it probably can’t do? It probably can’t determine that the picture I uploaded is actually a Ford not a Chevy, and no it doesn’t know what you mean when you say “the breaks kind of work”.
Most likely, the engineers will build it with pre-canned inputs and outputs, giving some room for custom natural language(but don’t stray too far out of bounds or you’ll trip their guardrails). This guided UX experience is now the antithesis of what makes the technology so fascinating. They would essentially be saying, “hey come use our fancy, creative AI tool. You can type in natural language how YOU want….well no, not exactly how you want, just use our pre-canned inputs and it’ll work”
Which leads us to the experience problem again. Who likes engaging with chat agents(human or AI)? No one. It’s a process we do out of necessity, and it’s not one we want to get creative with.
If we accept this as a ground truth, that no one likes talking to agents. Why are so many companies, and engineers, building chat agents?
It’s More Black and White Than you Think
The simple answer, is that it fits the technology the best. Large Language Models, are highly capable typers(and soon to be speakers). The best way to showoff your adept ability at understanding what another person is trying to say is to talk back to them.
The secondary answer, is that customer service help is the best way to keep your customer comfortable and happy. In a study, PSFT reports 74% of internet users prefer interacting with chatbots when seeking answers to straightforward questions. There is a transactional nature, not a dreaming one. Get in. Get your answer. Get out.
Got a question about your purchase or reservation? Need help understanding our confusing return policy? Want to start a return? Simply give it some information and you’re on your way. Often times this is infinitely faster, and more readily available than traditional human based flow.
The cherry on top? Your organization probably already has the data you need and the man power to build a customer service solution. You don’t need a fancy UX. You just need accuracy and relevant documents, which then feed an Agent generating high quality black and white inputs and outputs. Get the customer what they are asking for as efficiently and accurately as possible.
That’s what they want. Use the power of natural language to augment and improve your existing human interfaces. Your customers will understand it more. They will be prepared to interact with it, they’ve known about autonomous customer service bots for decades. It’s up to you and your team how you best serve them.