Chat bots need to be trained, but where do you find a lot of good data to train them with? Turns out, it could be right under your nose. Here’s how we’ve solved challenges in that space.
Gmail and WhatsApp Liaison Bot
Challenge: A building supply company needed a bot that sat between their Gmail and WhatsApp accounts. They were manually reading through certain emails that contained purchase orders, and then creating a WhatsApp group with all the necessary brokers, dispatchers and other contacts that were involved for their client’s address.
Solution: We wrote a bot that did exactly that. Client purchased a single 20-hour block and in that time, a fully configurable bot was created that would periodically watch their Gmail account, find key pieces of info out of certain emails, and build a group in WhatsApp based on the contact info found in that email, along with an Excel spreadsheet holding extra contacts that the customer wanted added for each of their clients.
Result: All of the above got done well under 20 hours, and it’s saving them hours of time PER DAY. They’re excited to find more things to work with us on.
Call Center Recordings as Chat Bot Training Data
Challenge: A client reached out with a sticky challenge: one of their own clients expressed a desire to train their online chat bots to answer questions before having to engage their internal call center. There were too many people spending too much time answering questions that could’ve been handled earlier. But, where were they going to get a massive amount of data to train the bot?
Solution: The answer: Google Cloud Platform. The call center at this client already recorded every conversation, and so the solution was to set up an automated way for GCP to take in the sound files every evening, transcribe them down into text, and put the transcriptions in a place where their system would scoop up the transcripts and load them into the chat bot trainer. As a bonus: everything got cleaned up so that files wouldn’t pile up for months.
Result: An amazing way to use the data that otherwise would’ve been sitting around, and repurposing it to save a ton of time later. The client didn’t end up paying a lot for this project but it likely saved hundreds of hours per month in call center time alone.