I used to be a chatbot. Obviously not the artificial intelligence kind you see popping up on websites all over, but that was effectively my function.
My first real job out of college was answering the 1-800-2-BANYAN marketing line for Banyan Systems a network operating systems vendor. Before there were websites, people used to call 1-800 numbers to learn more about a company. Whenever Banyan ran a marketing program — magazine ad, direct mail piece, trade show flyer etc. — the call to action was literally for the customer to call 1-800-2-BANYAN. And I had to answer that line, along with three other people.
We answered questions like “Can you help me set up email at my company?” and “If I have 2 offices will you’re software let me share documents without having to fax them or print and mail them?” Lots of rudimentary stuff, but it was early days of computer networking. The Internet “arrived” about a year into my time with Banyan but I answered that phone for another two years.
Another thing we did on that line was answer questions about model numbers and licensing. Existing customers would call when they needed to increase the users on their network or add a piece of functionality. We became experts at licensing even though that wasn’t our primary role.
After I had been there about 6 months I was invited to an operations meeting. The senior person on our team collected all the data about call volumes, talk times, leads passed and a few other typical call center statistics. She asked to go first and presented to management. Obviously we were very busy and doing a wonderful job. In fact we couldn’t really stay for the whole meeting because we were needed on the phones. I was young, new and clueless so I left with her, pleased that my job felt secure.
A quarter later I was able to go back to the operations meeting. I was a little more comfortable with what I knew and how things were working. I was also ambitious so I had a question — “Are we doing anything to simplify licensing? Lots of customers call about licensing and they’re pretty frustrated.”
This wasn’t the data they were used to getting from the kids working the 1-800 line (and it probably wasn’t appropriate for an operations meeting but that’s a different issue). Licensing was a problem and they were working to address it. I agreed to sit down with the product team and explain some of the customer sentiment we were hearing. Our feedback and customer conversations turned out to be very helpful in creating new license options.
While it may seem like a bit of a jump, this is the difference between a chatbot and an auto-responder. The chatbot allows for natural language processing and has an idea of intent. Reporting doesn’t need to be purely numerical — times engaged, length of engagement, branches on the decision tree, next hop, etc. — it can cover keyword frequency, related terms, positive and negative words and more. You get information about customer sentiment without needing a human to filter it.
And a chatbot can do this 24/7 with no call queues.
Like any technology chatbots are not flawless. They’re still evolving and getting better. But just like when I was the young employee learning the ropes on a 1-800 line a chatbot will get there and soon provide you with customer details you didn’t used to get.
On top of all this, they are supposedly “easy” to implement.
I’ve never coded for my paycheck, but I did do quite a bit back in high school and college. I was lucky that I got to start coding with C (even though people were working for Fortran and Pascal) and as one computer science professor put it, the language is fungible, it’s the concepts that matter.
As I continue to work at relating my interest in #AI to my work writing and selling science fiction books I’m going to try and implement a chatbot on my site. Hopefully it will become sort of a student becomes the master moment for me. though I suspect that it will be more of the student realizing they still have a lot to learn.
