The challenge
The customer organises meetings for its clients. Rather than clients contacting each other directly, all communication runs centrally through the customer, and it wanted to automate this process while letting users keep writing in natural language.
The customer initially used Google DialogFlow, but its performance was not good enough to automate the process fully.
Our solution
We delivered a solution that processes users’ replies to a meeting offer. Users can accept the meeting, decline it or propose an alternative date or time, and the system also handles automatic out-of-office replies.
From each free-text message, the solution extracts the information the scheduling process needs.
- Intent recognition: whether the user accepts or declines the meeting, proposes a new time or says the time does not suit them
- Extraction of a specific proposed date and time from the raw text
- Detection of out-of-office replies and extraction of the return date, if given
The outcome
The solution allowed the client to automate its scheduling process and reduce the number of human operators, saving 3,000 man-hours per month.
It also replaced a costly third-party service. The customer expects a tenfold increase in its number of clients, and the cost of the Google solution would have risen tenfold with it.
