I used to preach that it's better to have zero bots than one that knows next to nothing. I was mostly wrong.
Rather than have no bot at all, it's better to have a bot that knows two things: 1) How to get the user to enter some info, and 2) How to collect that info in a database and potentially pass it along to someone who can get back to the user.
The "smart" part comes later. How much later? Days, not months!
Day 1: Add the Botcopy chatbot messenger to your website.
It knows nothing except how to pass information to your team via a human handoff, ticketing, or @support email and store the queries in your analytics. Congrats, that's plenty for day one.
Day 2: Have a way of looking at the queries that came in yesterday.
Then, if you see a pattern, automate a response for that kind of query in Google CloudDialogflow, so it can send at least a few end-users off on their merry way, euphoric with instant gratification, thinking warm fuzzy thoughts about your brand. Perhaps they're thinking, "Wow, the IRS gets a bad rap, but that experience was fantastic!"
Day 3: Now that your bot is churning along, gathering data; it's time for you to step up and accelerate your bot's power.
How? Easy: Use your common sense and intuition to create some skills you know the bot should have. Now you're training your bot from not one but two data streams! 1) the queries coming in through your bot and 2) the stuff you know the bot needs to know because you have common sense.
Day 4: Keep training based on these two approaches.
Day 5 :Based on the top three most common incoming queries, add some suggestion chips to increase the odds of a correct match.
Users can still use natural language, but now they have some guardrails, too. Rinse and repeat, and your AI-first messenger solution will get exponentially smarter. Keep adding info, based on user demand, and based on what you want to automate.
The takeaway
As a result, the live agent and ticketing burden will eventually vanish or reduce to a fraction of what it once was. Your messenger, live chat, analytics, and cloud NLU should all snap into place nicely, a power stack that grows in any direction. In addition, your cloud provider is omni-channel and brimming with tools and services to connect your bot with your data and automate more processes. Impatience is a virtue. The sooner you start, the sooner the smart. Begin it.
