How a chatbot should be used, saving lives.
The emergence of chatbots has brought forth a wave of opportunities and possibilities that would make any technologist stoked. But sometimes the inner child in all of us can get a little too carried away. We begin to jot down things we want our next chatbot to do, most of them being things that are nice to have but not necessary at all.
Regardless of the resources and technical capabilities that one has at their disposal, problem-solving should still always come first. More importantly, simple yet effective solutions — the best problem solvers know how to KISS (keep it simple, stupid).
The chatbot that rose from the weeklong hackfest that Microsoft and Englander Institute of Precision Medicine (IPM) at Weill Cornell Medicine participated in, was a testament to that.
Weill Cornell Medicine chatbot
The knowledge base at Cornell currently supports 163 genes and 518 variants with 404 clinical interpretations. In order for clinicians to access medical data, they would have to click through a portal – clearly, a tedious and old method that could be automated, enter chatbots.
Microsoft and IPM spent a week hacking together a chatbot that could support both text and voice interactions with the knowledge base at Cornell Medicine.
Rather than clicking through a portal for information, users could query the chatbot for something in the knowledge base and get a response with the relevant data. This chatbot was integrated with an array of channels such as Microsoft Teams, Skype, Slack, and WebChat.
The chatbot enabled clinicians to access data in a handful of different ways, make clinical decisions at a faster rate and save time that could be better spent on their patients.
The technical delivery catered for both basic and complex scenarios
“Tell me more about EGFR.”
“Tell me about KRAS G12D.”
“Are there any clinical trials about this?”
Users would be able to query the knowledge base further with more complex questions.
You can the try the bot out here
Use chatbots in a meaningful way
The folks at Microsoft and IPM weren’t looking to create a chatbot that could jump over hoops and tell you your future, they built a tool that could automate a process and could scale.
They didn’t create a cool chatbot, they solved a problem in a cool way.
Chatbots are enablers for scalability, automation and an environment where users can focus their time and efforts towards more meaningful tasks.
If this read has left you inspired and you’ve identified a process that could be automated in your own operation, we’d love to hear from you!
Chatbot Automation Nirvana for your company in a five step strategy
Do you want more sales, more leads and happier customers? Less churn and more satisfying support requests? Don’t have the money for staff? That’s fine. Chatbot automation is your knight in shining armour that will allow you to scale one-on-one conversations with your customers. Without breaking a sweat.
This strategy outlines how the companies we partner with are automating their sales and support with a friendly, helpful chatbot. This strategy gets them to automation nirvana in a cost-effective way. No useless dollars spent, maximum learnings applied and happiest customer result.
What’s your secret chatbot automation strategy?
Alright, it’s really not that complicated. Companies we work with are always hesitant to go balls-deep, so to speak, investing in a chatbot. These concerns are totally warranted. What if chatbot automation isn’t all it’s cracked up to be? What if we sink thousands into a smart assistant and nobody uses it?
That’s where the strategy comes in. We at Mayte never, ever recommend releasing a full-featured project or product in one go. We always break it down into phases that can be easily learnt from and improved upon. This keeps investment conservative whilst producing the best results. Sort of like how you don’t want to jump the mark when designing a car.
But my business needs X and Y, STAT!
It’s also important to remember that businesses have priorities. These boil down to a few simple things: more sales/leads/customers and decreased churn via improved customer experience. Any innovative idea or project is always driven by a business or market lead so we can’t break the mould here. We need to increase leads and sales whilst making the customer happier. Chatbot automation can do all of this with the added benefit of long-term cost savings.
So here it is (I’ll go into more detail below). Here are the five steps we pitch and recommend companies go through on their journey to business and chatbot automation nirvana:
- Start with a Messenger bot and associated Ad campaign on a siloed Facebook page
- Add the Chatbot to your website after making improvements based on data collected
- Integrate your knowledge base with the chatbot to alleviate your support staff load
- Design and build common support tasks that are easy to automate and high frequency
- Integrate the resulting Assistant with your call centre.
Why chatbot automation in this order?
Remember the business priorities I mentioned? More sales, better customer experience and cost-saving. These five steps have been carefully crafted to hit the cheaper, low-hanging fruit first whilst also building a longer-term strategy. The cheapest and high impact (sales) initiatives are up front, whilst the longer-term, high cost and incredibly high impact ones are in the back. Business up front, party in the back. Nice.
Jokes aside, let’s break down each step of this chatbot automation strategy.
Baby steps: Messenger campaign
Chatbot automation became ubiquitous in the western world after Facebook launched the Messenger platform. They’ve grown the platform and added features non-stop. It’s now a robust tool that businesses can use to scale interactions with their customers in new and engaging ways.
The first stage involves creating a Messenger chatbot promoted through a Facebook Ad on a Facebook Page separate to your company page. There are great tools out there like Chatfuel and ManyChat that make it easy to get a messenger bot live (however can be restrictive in the future). They are a great place to start. Here’s why a Messenger campaign is a good first step.
Depending on how much you invest in branding and personality and what you want your chatbot to do, it’s a cheap place to start. There are plenty of things your chatbot can automate in this first stage. It’s important to pick and prioritise what will have the most impact on your company. It could be one of:
- Lead generation
- Lead qualification
- Promoting a new product
- Games and quizzes
- Dripped content
Regardless of what that low-hanging fruit looks like for you, the important thing is that you start on Messenger. Why? Two reasons. Firstly, it’s feature-set is incredibly diverse. Secondly, Messenger as an Ad destination.
Out of all the platforms we’ve worked with, Messenger has invested, by far, the most in the UX of using a chatbot. The quick replies, buttons, cards and interactions make bots without NLP incredibly engaging and also open up new opportunities as developers to do incredible things.
Messenger as an Ad destination is the real game changer for us though. When setting up an ad you can specify that the user gets directed to a Messenger conversation instead of a landing page or a regular Facebook page. This is HUGE because it means the entire interaction happens where users are already spending their time, without leaving. They see the ad, interact and convert all within one of the most used apps in the world.
All of this is done through a siloed, branded Facebook page. This is usually to allow the main page to continue to run as not to disrupt the day-to-day ongoings of business. It also makes stakeholders freak out less! Worst case the campaign goes bad the page can be chopped off at the legs and there’s no association to the main, original facebook page. That generally only happens if you don’t design your chatbot correctly, but that’s another article.
Finally – be sure to track your bot using analytics! This will provide valuable insight into how your conversation could be optimised. Chatbase is a great tool by Google for this.
Congratulations – you’ve now opened another channel for sales and leads.
Stand on your own: Add the bot to your website
Your chatbot automation campaign on Messenger should answer an important question: has this exercise benefited the company? You should be able to deduct from your metrics whether you’ve seen an uptick in sales, received more leads or handled an overall higher volume of traffic conversations with your customers in a positive way.
Conversations with your customers – this is the important takeaway. Traditionally scaling one-on-one conversations with your customers is a costly endeavour. Not anymore.
It’s time to bring your new secret weapon to ALL your customers. Launch the same chatbot on your website and as the first line of defence on your full-time facebook page. There are a few ways to do this. Facebook recently released customer chat on your website – so you could just deploy it and be done with it. That’s the easy version.
This is a great time to integrate the successful bot with your existing customer support tools. Unfortunately, support for rich interaction like quick replies and cards is currently limited, however, there are workarounds.
We recently integrated Intercom (supports only images and text at the time of writing) with a messenger bot and faced this challenge (bot was built with meya.ai). Instead of integrating the bot through Intercom, we integrated Intercom with both Messenger and our custom chatbot. This has the benefit of having a running conversation with your customers AND handling hand-over.
Hand over is a topic in itself. I will say that it’s extremely important to allow for this in some way, otherwise, you’ll have a handful of poor experiences. People get easily frustrated when they cannot achieve their goals quickly and this is the entire purpose of having a chatbot. Handover can happen in many ways – listening for a language trigger – i.e. “I want to talk to a real person” or tapping a menu item. Be sure to make the transition flawless, else risk a poor experience.
The bot now acts as a ‘first line of defence’ before your customer support team. Not in the sense that it tries to deflect the problem, but tries to address it automatically before deferring to a human.
Get walking: Boost self-service capabilities and save time
One thing has been consistent with the companies we’ve worked with in automating the customer experience: thousands of hours wasted on menial tasks and questions. Think about it – when you call up a customer service rep, there’s a micro onboarding. Privacy statements, verification, even waiting in a queue. All BEFORE the crux of the problem is actually solved.
You’ll see the numbers if you research and look closely enough – but around 80% of your requests are things that customers can generally do via self-service. Self-service has been a major paint point since the dawn of capitalism. How do we give our customers the tools to leave us alone and make more money?
Interview your support staff and research what are the most common queries and problems. Sort these by the ability to be done without your intervention and you’ll end up with something that looks like FAQs.
Shortcut: just take your existing FAQs or existing knowledge base.
There are plenty of tools to do this quickly and easily. QnA Maker by Microsoft is great. If you’re feeling adventurous, it’s time to dive into Dialogflow, Amazon Lex or Wit.ai.
Be sure to revisit and re-write the answers that your chatbot spews out. Chatbot conversation design is important to get a great experience.
Implementing this will alleviate your staff by giving customers answers faster, more consistent responses. It bolsters the ‘first line of defence’ from stupid questions and allows your staff to focus on the curly, more complex problems your staff face.
Start running: automate the hard stuff
By this stage, you’ll have a chatbot relatively capable of understanding customer requests and processing menial tasks. That’s all well and good – but as it only throws answers back to frequently asked questions, it doesn’t actually do anything core to the business.
This phase can be the most challenging, as you really have to balance what customers do most with the cost of the build. Complex features can take a lot of time and resources. Looking back to the effort/impact scale can really help here.
There are some great examples of companies doing this already, shown below.
Compete in the mother f*cking Olympics: automate your call centre
When you’ve successfully automated complex business tasks via conversation, what’s next? Provided you used Amazon Lex or similar for your Natural Language Processing, you can plug into tools like Amazon Connect to directly migrate these tasks to a voice interface. Check out the video below to have your mind blown.
So there you have it. Our strategy for free. Go forth and make bots! If you think your customer experience can be improved via automation, feel free to drop us a note!