Resolve recurring requests automatically, step by step

In the Flow Builder you put a conversation together from building blocks by drag and drop. A message, a question, a few buttons to choose from, and an action at the end, such as “Mark as resolved”. Nobody has to write any code.

A flow starts either with the greeting or as soon as a particular keyword comes up. If someone types “return”, the returns process starts. Questions outside a flow are still answered freely from your documents. The two work hand in hand.

When the flow asks for an email address, it only accepts a valid one. The same goes for phone numbers, numbers and dates. If the customer types “max.mustermann.com”, the bot politely asks for a correction. The typo is caught before it reaches your system.

Depending on the answer, the flow branches. A warranty question goes one way, an invoice question another; a device under two years old one way, an older one another. The bot saves the answers straight to the contact profile. If a person takes over later, all the details are there.

What the flow collects often needs to go somewhere else: your ticketing system, your CRM or an automation of your own. That's what webhooks and the REST API are for. You'll find the API documentation in the dashboard under “API token”; API access is included from the Professional plan.

We don't promise a ready-made integration with any particular ticketing system here. Whether and how your system can accept the data is something we're happy to look at with you beforehand.

The analytics show conversation volume, how many conversations the bot closed on its own and how often it had to hand over to your team. On top of that you see the most frequent questions and the ones it couldn't answer. Advanced analytics with conversation histories come with the Professional plan.

We use this ourselves for supporting our own software. There, the bot answers 2,100 questions a month automatically, first-level tickets are down by 47 %, and the team saves 9 hours of support time a week. Most of the work, by the way, wasn't in the bot. It was in filling the gaps in our documentation that the bot showed us.