It's the age of the chatbot. The chatbot revolution is coming. Unless it isn't.
Conversational User Interfaces are still very much on the Innovation Trigger rising curve of Gartner's hype cycle. And the people who are the most hyped up about chatbots seem to be the people that sell platforms to build them on. As chatbots seep out of the consumer space and into our internal workplace tools, are we at risk of losing our job because a bot can do it better?
Back in April 2016, Facebook opened up the world of chatbots to businesses and brands. Zuckerberg was keen to share the vision of people conversing via Facebook Messenger to an intelligent bot, as if it was a friend. We’d be booking hotels, sending flowers and making restaurant reservations without needing to go near a website or an app. Hooray!
Fast forward to the present day and consumer chatbot adoption has been hit and miss. Despite more than 100,000 bots living on the Facebook Messenger platform, I interact with zero on a regular basis. One retailer in my area has abandoned their bot in favour in a new ordering app. The Singapore Government has a bot to ask about pollution levels or send feedback about their services, so maybe I'm just in the wrong market or the wrong demographic. As a technology lover, I feel like I should be a chatbot fan, but as a consumer, they just don’t do it for me.
Back in the office, bots are the new black
It's all your fault, Slack. That easily deployed platform comes with more than 600 bots and app integrations, waiting to be turned on with the flick of a switch. They promise connectivity, productivity and just plain fun. The IT team is monitoring uptime and SSL certificates, JIRA issues, PagerDuty incidents and Visual Studio Team Services projects all within the comfort of the Slack app. This made bots a must-have for Microsoft’s Teams app. It included some popular ones like Polly for polls and Meekan for meeting scheduling, then threw in AzureBot for managing Azure VMs. Like Slack, Teams also uses a bot interface for its own internal help search.
The possibility gap
Enterprise internal use of bots is apparently the way of the future. Like the consumer bot hype, you'll find lots of examples about how bots could be used in large organisations but few actual case studies on bot successes. Apart from the chat-based platforms above, bots may be added as part of an application's functionality or exist in a small proof of concept with a limited number of beta testers. However, no one is saying they've revolutionised their IT help desk processes or change management approvals because of a chatbot.
The usual suspects
Gartner has offered some potentially good ideas regarding chatbots in the enterprise. Its top suggestion is reducing the call centre help desk workload by answering FAQs and routing requests to the right queues, complete with an automatic update back to the end user. These kinds of customer service tasks are most commonly top of the bot capability list, but you have to wonder if end users are ready for it. If they are already happy emailing the help desk instead of phoning, how likely is it that are we'll lure them away from email in favour of a chat interface?
Take APIs and a chat interface and you have the next step in ChatOps automation. GitHub supports this world view, releasing its open-source Hubot to the public. Who'd use a GUI or even a script when conversing with a bot makes your work visible to your entire team? Add in some approval steps and we have digital change management. If your IT team has moved from email to chat as its primary communication platform, this use of bots makes sense. That's a big assumption to make, though. Shifting to a chat-based way of working is possible (you can argue if it's preferable) if your organisation cuts through the shackles of historical work processes and rebuilds them for a chat-centric world. It's the cool thing to do, though, whichever chat platform you choose and is more likely to be led by the IT team as the first adopters.
Gartner also mentions bots as scheduling agents. Given that email meeting tennis is boring, there's quite a bit of activity in this space already. Meekan is the bot for the job on Slack, Microsoft Teams or Hipchat but does require one of these platforms to run on. It supports Google Calendar, Office 365 and iCloud. Outside of chat, X.AI automates the emails backwards and forwards by pretending to be either Amy or Andrew. The paid business edition can pose as your own domain name instead of @x.ai but only if you use GSuite, with custom domain support for Office 365 "coming soon". X.AI's biggest limitation right now is it only supports a maximum of four meeting participants, but they plan to increase that. Meanwhile, Microsoft is training Cortana for the role with the Calendar.Help project in preview. Cortana will check Office 365, Outlook.com and Google calendar and can also schedule Skype for Business meetings. One cool thing about these scheduling bots is that you can specify a time range in your initial request, even if other times in your calendar are technically free, to give some guidance to your own meeting time preference.
What we want, what we really, really want
It's easy to expand on IT components like monitoring, alerting or querying IT systems. Give me a bot I can throw natural language queries at like "How many people are in the XYZ security group?" or "What's the current memory utilisation on Server G?" and I'd be a happy (and lazy) IT person. Step away from IT and what does your organisation actually do (assuming you're not an IT organisation)? Will healthcare bots let doctors ask questions about historical cases or infection trends? Will paralegal bots eventually end the need to trawl through dusty case files, as an intelligent bot surfaces relevant historical rulings? Will police bots link seemingly unrelated data to help connect contacts and solve crimes? This all sounds futuristic but it's more exciting than a bot ordering my pizza.
So, how do you get past the hype?
First, you have to decide if chatbots are a priority project. The reality of enterprise IT is that there are always fires to fight. Right now, we've got the business screaming at us about digital transformation and disruption, because our competitors will eat us for breakfast if we don't do something. That's maybe why consumer-facing bots are more prevalent than the use of bots for in-house capabilities. How much of a priority is a new chatbot compared to the rest of your transformational projects?
Next, you'll need to pick your platform. If you're using a chat-based collaboration platform, that will influence what you use to build your bot, though there are a huge number of choices. Will you build your own using something like Hubot, IBM's Watson or Microsoft's Bot Framework, calling on Microsoft's Cognitive Services? Will you look to a third party like ChattyPeople, Smooch or FlowXO? Look at where the data is that you want the bot to read from, interact with or change. Last thing we want is a data security nightmare on our hands.
Finally, be prepared to invest in training your bot. Users are not going to be happy with "I'm sorry, I don't understand." Services like LUIS (Microsoft's Language Understanding Intelligence Service) can help, but English is still a tricky language and people will ask for the same thing in different ways.
You can replace the human, but you can't replace the humanity. Bots give us the capability to automate routine tasks or seek information with a conversational interface. It's a doorway to replacing some human roles, though we don't seem in a hurry to totally embrace it. There will still be a need to talk to someone, when a subject requires empathy, or making an exception. Bots aren't good at handling things outside of the normal. That's going to require better artificial intelligence, which we are simultaneously working towards and are totally scared of.
It's going to be interesting to see if the IT workers of the future will step into a bot-powered workplace or if this will all fall away as hype. It is likely that we'll see further automation in the enterprise, but will we give up other methods of working in favour of natural language interfaces? The jury's out for now. ®