Step up your automation game with AI

Training and testing your AI Agent

Now that you have a good dataset for your AI Agent, it’s time for your AI Agent to start learning this new knowledge. We do this by training the AI Agent.

In Watermelon, training your AI Agent is a fully automated process in which the AI Agent analyses all the new datasets. The AI Agent will look at the different variations on the questions and calculate all the other possible variations. By training the AI Agent, you ensure that your AI Agent will be able to recognise questions, but also recognise spelling mistakes and completely different structures of sentences.

After you have trained your AI Agent it’s time for testing. Testing your AI Agent is a crucial part because in this phase you can see whether your dataset was actually good or not and where you can improve your AI Agent. But before you start the testing, it’s important to decide on a testing method.

Internal test group

The easiest way to test your AI Agent is by setting up an internal test group. This can be a group of colleagues, preferably from different departments to maximise the amount of variety in feedback. The benefit of an internal test group is that your colleagues have a good understanding of your business and know exactly which topics to try out in your AI Agent. The downside of testing internally is that your colleagues probably already know which information is available in the AI Agent and are unlikely to ask the same questions as your customers.

External test group

Instead of asking your colleagues to test your AI Agent, you could also look outside of your company and ask a select group of customers to provide you with their feedback. By setting up an external test group, you can get a lot of valuable feedback because your customers don’t know which questions and variations you have added to your AI Agent. They will naturally ask their questions in any number of ways, providing you with many new variations you can add to your datasets.

Whatever your preferred method of testing is, make sure to take your time with this phase as you don’t want to rush it. Gathering and analysing feedback is a valuable way to quickly optimise and expand the capabilities of your AI Agent.

Once you have tested your AI Agent completely, you are ready for launch. Make sure that you have added all the necessary knowledge, your datasets are complete and your AI Agent is activated on all your channels. The last thing to do is to press the big, green button and you have your very own, fully functioning AI Agent!