Contact us
Kontakt

Conversational AI: From preschool to graduation

5 min 4. 10. 2022

A decade or so ago, every website had a chatbot. You were nobody if you didn’t have a pop-up window with an annoying beep on your homepage. In short, it was a trend, and in the vast majority of cases the bots were simple solutions with a few programmed scripts, which they could not deviate from. Now, a few years later, the situation is different.

The original button-based “preschoolers” are, it’s true, here to stay. And slightly smarter – but still of limited use. However, they’re now joined by more educated colleagues, who are bringing artificial intelligence, machine learning and NLP technology to the table.

The preschoolers might be enough for you. Just don’t expect miracles!

Let’s build up an analogy with the education system here. We’ll start with “preschoolers”. These are the simple button-operated chatbots that can handle selected simple tasks. Examples include:

  • Information on opening hours
  • Welcoming users to the site
  • Simple surveys

You can see for yourself that this is all useful advice.But it takes place in a strictly scripted way. If you ask a preschooler chatbot anything unexpected, it won’t be able to advise you. And then it will either lock up or offer to contact an operator.

blog-img

Elementary and secondary school students

Purchasing assistants are slightly higher up the educational ladder. They help you choose a product (maybe goods or maybe a service) that is tailored to your needs and will fundamentally improve your customer experience. They’ll guide you through a set of questions and direct you to the answers – or connect you to an operator again. To make them work right, generally all you need are correctly filled-in product sheets and dozens of conversations from chats or emails. But here again, we crash into the fact that once you deviate from the script, a simple chatbot can’t cope.

If you have hundreds of conversations (it doesn’t matter if it’s chat or emails) that a chatbot can learn from, you can create an FAQ bot. It recognises specific words in the FAQs and is able to answer them correctly. But don’t expect any great freedom in conversations from these bots either.

Education continues at university

Machine learning, conversational AI, NLP. Just as a whole new world opens up for you as a human at university, a chatbot or voicebot can do so much more when it’s learn advanced technology. Because even if you have cute little conversation trees ready, users are often unexpectedly creative in asking questions, and then traditional solutions are no longer enough. That’s why it’s important that the chatbot really understands the person, which is where NLP (natural language processing) comes into play.

But... While English, for example, is a relatively formal language and thus easy to process, Czech is “worse” in this respect. Therefore it is often better to develop your own solutions for Czech than to rely on foreign NLP from a large company. When users delight in a rich vocabulary, interesting turns of phrase, inflections or adjectives, that’s where the developers’ nightmare begins.

Along with their own NLP, Coworkers also uses a neural network and machine learning. Thanks to this, specific words are interlinked, and the intelligence behind the chatbot infers a site visitor’s intent. Or the caller, because the same system works for voicebots. Meanwhile, the bot is always learning, and with each new conversation, it either reinforces the correct answer or adds a new option to its memory. And when the time is right, there are still buttons. These can sometimes complement advanced solutions.

blog-img

Who will your new colleague become?

It’s only logical that each chatbot or voicebot is best suited to something different. For small online shops, simple websites, or projects with a few hundred visitors, a basic rule-based solution will often suffice. The retailer tries out a new and interesting form of communication, does something extra for their visitors, and gains a colleague who can solve simple questions quickly and by itself. They don’t need advanced solutions – they can manage their business on their own or in cooperation with a team.

But the real magic happens when you start getting more and more orders and your visitors number in the tens of thousands. The technologies mentioned above can handle up to 70% of routine conversations. If we’re looking at, for example, on an insurance company, an AI can handle situations such as:

  • Insurance claim reporting
  • Insurance payments
  • Client authentication and login

And meanwhile the client doesn’t even need to know they’re not talking to a real person. All in all, this saves a wealth of time, which operators can then devote to non-standard situations and requests. This is also one of the most important ideas behind Coworkers: chatbots and voicebots aren’t meant to replace humans. On the contrary, they complement them perfectly.