Build an emergency help bot with Coronavirus best practices in minutes

Matt Wade
AtBot
Published in
7 min readMar 20, 2020

--

Note: the AtBot team is eager to help those organizations with a real need for sharing good, timely information to your stakeholders regarding COVID-19 status in your area. Please do not hesitate to reach out to us for assistance and guidance to get you going immediately. Email hello [at] atbot [dot] io now.

The Coronavirus pandemic is proving a real need for automated self service tools amid millions of people wanting information and reassurance about COVID-19 status in their area.

Overwhelmed customer service teams, health departments, and hospitals everywhere are unable to deal with the influx of requests coming into them. Chatbots are an easy, affordable, and scaleable option for handling these tasks in a way no human or standard IT system can.

And AtBot offers even more value than the typical chatbot because its low-code foundation means you can have an emergency response bot up and running in no time.

We think AtBot can be a means to sharing critical information with the public, your employees, customers, and partners, in a quick and easy way. Bots are available 24 hours a day, they don’t get sick, and they’re inexpensive to build and operate.

We decided to put together a proof of concept to show you the power of a chatbot combined with information and data that’s freely available on the internet. We built it to be a texting bot: you would text a phone number with your questions and get responses over SMS. But you could deploy the same bot to a website, mobile app, you name it.

This project is not one that’s meant to be reproduced exactly. This is a demonstration only and does not represent its data sources (especially the US Centers for Disease Control and Prevention); we simply want to explain and demonstrate the the ROI that comes with a smart chatbot in these uncertain times and hopefully provide some inspiration for how you can put one of these easy-to-build solutions to work for you, now.

In only a couple hours, we built a text bot that answers FAQs with CDC answers and pulls status reports from all states and the US as a whole.

There’s a lot of information out there about COVID-19 and how the public should be responding to it, from following social-distancing routines to checking themselves for symptoms.

In only a few hours, we were able to put together a bot on the AtBot platform that uses CDC-sourced frequently asked questions combined with reports on statewide reported cases of COVID-19 within the United States. This is only the beginning.

The CDC itself recently published its own bot on its website to answer common questions about COVID-19 and you can do the same thing plus more with AtBot and its underlying AI technologies to help ease the burden of sharing accurate, useful, and actionable information with your stakeholders.

The CDC recently released Clara to help with COVID-19 questions. You can easily do something similar.

FAQs and QnA Maker

Our AtBot-built proof of concept was able to scrape the FAQs from the CDC’s website using a QnA Maker knowledge base (KB). This process takes literally seconds. Using an AtBot bot, you can connect this KB and your bot will immediately start reacting to questions by sourcing answers from the KB and responding with what it thinks is the most appropriate answer.

This gives you the opportunity to review all the questions and answers CDC has provided and 1) remove some that aren’t relevant to your situation; 2) add further phrasings for questions to cover more ways someone could ask that question; and 3) add more question-and-answer pairs that are appropriate to your stakeholders. Basically, you’re seeding a knowledgeable bot with everything it needs as a foundation, plus configured knowledge that your team needs it to share with your stakeholders.

For example, maybe you’re a county health department and you want to answer the simple question, “How do I contact you?” You can do that. Or maybe you want people to know whether you have testing kits. You can include that in your own QnA pair as well. The opportunities for sharing good, timely information are endless with your COVID-19 chatbot.

You can scrape the CDC’s FAQ page into QnA Maker in seconds and add further phrasings and question-and-answer pairs to your liking.

Statewide status reports and Power Automate

We built our bot to be able not to just answer static questions, but also look up information from a dynamic set of data. In our case, we’re looking up the COVID-19 infection rates across the United States; that means each state and the country as a whole. Useful information, we think. That takes a little logic behind the scenes.

Power Automate is the built-in logic engine for AtBot, which gives you the ability to create logical steps, questions, requests for input, etc. without having to get heavy into custom coding — we’re not all programmers, right? If you have Office 365, you have Power Automate.

We found a useful GitHub repository that was constantly updating lists of Coronavirus cases, deaths, and recoveries called NovelCOVID by GitHub user dicedtomatoreal. The data is apparently coming from Johns Hopkins University, but it’s a random internet stranger who decided to put this all together. We thank dicedtomatoereal for doing this.

That said, we of course caution anyone using this data to confirm this resource’s accuracy before putting it to use.

But sometimes you just need to get stuff done. In this case, we set up a simple Power Automate flow to digest the report on a requested state and provide a human-friendly response via the bot. A combination of AtBot actions in Power Automate plus parsing some JSON gave us a concise, data-centric update about the spread of the virus within the US.

To ensure the bot knows which US state it was supposed to look up, we included some basic, but powerful, natural language processing. Using Azure’s Language Understanding Intelligent Service (LUIS), we can have the bot automatically understand what you’re asking and about which state.

LUIS already knows the names of all the states, so we just had to provide a few different ways to ask the question we’re answering with the workflow. Then we connected it to a Twilio account to support texting; we could also have added it to a website using a little JavaScript.

And below are screenshots showing what those tools look like.

AtBot brings together LUIS for language understanding and Power Automate for logic. In this case, a simple query of the state database can provide a powerful, timely response to someone texting a bot asking for COVID-19 updates.

Bringing LUIS, Power Automate, the GitHub API, and Twilio together through the AtBot platform, we have an inexpensive (likely less than 1¢ per question) and easily maintainable bot that even the folks in the health department can manage themselves, without any IT support.

And this is just the beginning. Many other actions and queries could be useful to your stakeholders. Maybe emergency room status in hospitals in your county is relavent. Track the status and provide up-to-date answers for people pinging your bot.

As you see your bot being used, track what people are asking and be sure to provide answers to those questions — whether they’re static answers from QnA Maker or are dynamic and will require Power Automate. Each additional skill your bot has is that many more questions you won’t have to answer over the phone or via email.

Giving our bot a voice

We originally set this bot up as a texting bot: once you have the phone number, you can text it any question you’d like and it will respond with the best answer it knows. Simple, but insanely effective. As mentioned earlier, we used the Twilio channel in Azure to make this happen.

But what if someone called the number? We decided to take our bot to the next level. Twilio also provides interactive voice response (IVR); you may know it better as the computer voice talking to me on the other end of the line. Using IVR, we were able to have the bot accept verbal questions. Your spoken questions run through Twilio’s speech-to-text service, that text is sent to the bot, and the bot comes back with the answer it otherwise would if the input was a text message. Then the response is read back using text-to-speech technology to you over the phone.

This is a super useful solution considering many of the people looking for information may not be SMS-savvy. Again, a simple connection through the AtBot platform made this quick and painless. Check the quick video of the bot responding below.

(Note: due to the way Twilio works, there is a slight delay after someone asks a question. You can minimize this through optimization.)

Wrap up

We think chatbots are a perfect fit for responding to this crisis by ensuring people stay informed and calm without having to wait for hours to get answers to simple questions and submit requests.

You can get started now by setting up an AtBot free trial and building your bot. And we can help you along the way. Reach out and we’ll provide the guidance you need to get your bot up and running immediately. Simply email us at hello [at] atbot [dot] io.

AtBot brings AI within reach

AtBot is the premiere bot-as-a-service solution for the Microsoft cloud. Built completely within Azure, AtBot is your out-of-the-box, easy-to-configure bot for Teams, SharePoint, or the web. Teach AtBot tasks using Power Automate, make him your corporate source of knowledge with QnA Maker, help him understand almost anything your colleagues could ask thanks to LUIS, and manage his features with the AtBot Admin Portal. Get going with AtBot Free or start your free trial of AtBot Premium today.

--

--

Matt Wade
AtBot
Writer for

Microsoft MVP • Office 365 & Microsoft Teams specialist • NY→USVI→DC→NY