Want to get started with a chatbot yourself without having to write a line of code? Thanks to generative AI and Google’s underlying PaLM 2 model, it’s ready in no time.

Artificial intelligence is everywhere today. With the launch of ChatGPT, we are talking about an AI hype that many organizations want to join. Anyone who isn’t there risks missing out. Generative AI can seem quite intimidating as a project, but fortunately there are many tools that can take you very far in just a few steps.

According to Tristan Van Thielen, EMEA ML and AI Tribe Lead at Devoteam G Cloud, integrating an intelligent chatbot is a simple first step towards integrating generative AI into the business.

All data is welcome

To make a chatbot useful, you need to feed it information. This used to be a time-consuming task that intimidated many organizations. Today this is no longer necessary as generative AI understands the context. “A chatbot no longer just needs structured data, which often takes a lot of time,” says Van Thielen.

He shows an example within Google Cloud where you can use generative AI. “As an organization, you have been able to use Vertex AI Search and Conversation for two months now. The PaLM 2 Large Language Model (LLM) is used, a powerful foundation trained by Google that is now also used by the AI ​​assistant Bard.”

According to him, it takes no more than an hour to register, collect data and insert it into your website or application.

PaLM 2 as LLM

An hour seemed short, but Van Thielen points to the power of generative AI, which can make sense of data to quickly train a chatbot. “Take your company website homepage, Ctrl A, Ctrl C, Ctrl V into a word processor, save it as a PDF, that’s all.”

This is so easy because PaLM 2 understands when the text says “About us”. The company information listed below will be forwarded to the chatbot. Since it understands the context, it saves you a lot of time.

The finer you represent the data, the better the result, but the model can handle general data quite well.

Tristan Van Thielen, EMEA ML and AI Tribe Lead at Devoteam G Cloud

The finer you present the data, the better the result, but he says the model can handle general data quite well. Another useful data set is your website’s FAQ. Copy, paste and your chatbot will be much smarter.

You can do everything we describe above yourself without having to write a line of code. At the end, Google will provide you with a code example that you can integrate into the website. Very important: All data remains your property. Google does not use your input to further train its own model. Anyone who uses the AI ​​assistant Bard today passes on their input to Google to further refine the AI ​​model.

Even smarter with Bard AI

You can also integrate the rest of your applications and data with this chatbot, but there’s more to it than that. “As soon as you want to put an entire product catalog in the chatbot, we have to work tailor-made because there is a lot more to it.” Take a hardware store, for example. When a customer asks for kitchen renovation ideas via the chatbot, PaLM can generate various examples thanks to the intelligent AI model. You can’t train something like that yourself, it’s far too complex.”

“By activating PaLM, the digital visitor receives a richer experience. For example, you can create a link to a website about kitchen cabinets after some kitchen results. For example, you helped the customer with a creative question and he or she suddenly sees a part of your web store to continue shopping. Or they ask some additional questions to refine their selection.”

Cost price

How much does a chatbot like this actually cost? That depends on how often it is used. You can find the prices transparently in the Google Cloud blog. Specifically, it depends on how many questions are asked. You can purchase $1,000 worth of credits. This means that 83,333 questions can be asked to the chatbot. If you go beyond that, it costs up to 0.2 dollar cents per 1,000 questions.

Interestingly, in addition to a chatbot, you can also activate the AI ​​model to record phone conversations.

Tristan Van Thielen, EMEA ML and AI Tribe Lead at Devoteam G Cloud

Interestingly, in addition to a chatbot, you can also activate the AI ​​model to record phone conversations. If the caller asks a simple question, such as opening hours or to find out whether someone is in the office, the AI ​​bot can answer or forward it to a “human” if necessary.

Look for additional interaction

For example, why do you need a chatbot on your website? According to Van Thielen, it offers (potential) customers an additional way to seek interaction. “Today we realize that too many companies want to do something with generative AI, but don’t know where to start.” A chatbot can intelligently support customers through your company’s offerings. This ensures that they have a better experience and can make a significant difference between the competition.

“Then a chatbot or a clever addition to your switchboard is a practical way to get started without much difficulty.”

—————————————-

This is an editorial production in collaboration with Devoteam G Cloud. The company is hosting its Gen AI For Business event on November 8, where you can build a chatbot yourself. Interested? Register here shortcut.