So you have a Chatbot. Are you ready for BotOps?

Today there are more than 200,000 Messenger Chatbots but 90% of them don’t use AI. They were built with simple, rule-based logic and predefined steps. Using AI to understand human intent is hard, but I don’t believe we should sit back and wait for technology to “catch up”. By creating new processes for the development and delivery of conversational experiences, we can find a path to more delightful automation.

In the Bot Stack, there are no shortage of tools to create a conversational interface and big players like Google, Facebook, Amazon, Microsoft and IBM are heavily invested in the underlying intelligence. Need a bot making platform? You can choose from more than 50 at this point ranging from no-code visual tools like Chatfuel and ManyChat, to developer frameworks like Microsoft Bot Framework and about a dozen others. I think what’s missing in the stack is a middle layer that better connects the top layer (Interface) and the bottom layer (AI), so that businesses can better monitor and manage their conversational experiences.

What is BotOps?

If DevOps helped teams build, test, and release more reliable software faster, BotOps brings an additional set of processes that specifically address the current and future needs of conversational software.

BotOps are development and delivery processes that aim to make conversational experiences more reliable and scalable.

BotOps can change the way you think about building a Chatbot. Much like DevOps, BotOps emphasizes better collaboration between engineering and operations. You’re building an experience for messaging which at it’s core is an interface for communication with your customers, so with BotOps, developing the conversational flow should be at the forefront, owned by stakeholders who communicate with customers — Product, sales, marketing, and customer service can define and manage this and be active contributors to building on this over time, while richer components can be engineered. For example, an “Appointment” function could be engineered, or integrated with a calendar backend, but it’s function in the conversation would be triggered by a sales team, who maintains dialog with the customer. BotOps enables to streams of development to operate in parallel:

I’ve seen people responsible for communicating with customers alienated from the artificial intelligence layer, while engineering resources are empowered with training tools that fall outside of their domain expertise. Why are bot training tools so developer-friendly, while UX and interface tools provide limited, or no accessibility to the intelligence layer?

BotOps seeks to make sophisticated AI more accessible to non-technical team members. Janis, for example connects in seconds to Dialogflow, and makes leading conversational AI from Google more accessible to non-technical team members. These non-technical team members can script and test conversations powered by AI all from the comfort of Slack through an intuitive interface and within their existing workflow.

When bot training with AI becomes more accessible through collaborative work spaces like Slack, you start to manage a Chatbot, not like a piece of software, but more like a member of your customer experience team, and you can monitor and even take over when necessary. Deeper functionality in the conversational experience can be delegated to engineers and Janis connects in minutes to many popular bot building platforms and frameworks that engineers and even marketers already use and love.

A better path to automation

Building a bot is a problem already solved. There are no shortage of options. I believe there will be two key forces that will help businesses achieve their automation goals over the next 10 years. The first will be driven by the big players who will continue to invest in, and advance the underlying AI technology that facilitates natural language processing. The second will be an operational layer that helps companies better manage progress towards their automation goals while retaining their users in the process. A Chatbot will only be as good as the training that goes into it and that takes time, but if you can manage and navigate your path to automation while retaining your users in the process, well that seems like a winning strategy for messaging in the long term.

Before you go, some things to consider:

1. What are your current operational processes for your Chatbot? What are your challenges and how are you solving them? Curious to hear in the comments.
2. We’re building an AI assistant for bot makers at Janis.ai to make BotOps easier. Check it out.
3. Recommend or share this if you found it useful. It gives me 🔋 to write knowing people find value in it.