Compare a web shop to a store – and conversational commerce with the salesman who guides you past the cash register. If you make that comparison, it immediately becomes clear that it’s of vital importance to start the conversation on the right moment. Sometimes you enter a store and just want to look around for a bit. A salesman who immediately rushes up to you will put you off. At another moment you can actually use some help and that shouldn’t take too long, or you’ll leave the store again.

Just like for a salesman of flesh and blood, it’s important for a bot to address the client in a friendly way. You expect answers of your salesman – so you expect the same of a bot. Therefore, it’s best that your bot knows beforehand what he can or cannot do. You better not place your bot in your shop while untrained. Bots – unlike real people – can expect a series of unexpected questions. For instance, ‘What’s your name?’ and ‘Am I talking to a real person’ are questions that are rarely asked to store personnel.

With a bot on your web shop you can sell much, much more than without. The bot can actively do up- and cross-sell suggestions, that were otherwise a bit ‘hidden’ on the product page.

The user experience plays a key role in the interaction with a bot. People expect a bot to be fast and intelligent. A tree structure with predefined questions and answers is not intelligent. This type of bot is easy to program, but will not be used. Bots are not to be mistaken with a new version of the telephone menu. The ideal combination seems a good ratio of free text (2/3) to buttons (1/3). When you let a client make a product choice, you can offer him or her a so-called carrousel of possibilities with attractive images and (for example) the possibility to click through to more information on the website. Of course a bot should never interrupt when someone is typing and he proactively informs the client when an answer will take some more time.

Just like during the development of a website, it’s a good idea to, when developing a bot, first develop a prototype of the most important conversation flows and ask users for feedback.



It’s necessary to arrange a seamless transition from bots to live chat. After all, clients expect an instant answer, and when the bot can’t offer an automated answer it will result in dissatisfaction. Practice shows that bots, despite their self-learning system, can’t answer everything yet. Such a chat could take over, reactively (on demand of the user) or proactively (on the basis of a quality measurement of the conversation).

When your bot is 24/7 at the service of your clients, then such an escalation is possibly difficult to arrange. After all, your employees won’t be available 24/7. You can then also opt for a hybrid approach with an external service desk.



(Fields marked with an asterisk (*) are mandatory.)