Make way for the bot colony
The year 2016 will be remembered in technology circles as the year that the bot craze reached maximum volume. Not surprisingly, Facebook had a big hand in building the momentum. Since the tech behemoth announced a bot developer framework and distribution platform in April, the market has been flooded with upstart companies (and stalwarts such as Microsoft and Oracle) looking to ride the wave. And what a wave it is: The artificial intelligence (AI) market — of which bots are a major subset — is expected to become a multibillion-dollar industry in the next few years, and VentureBeat published an overview of the bots market that found 170+ companies, $4 billion in funding, and thousands of bots currently in the space.
But as it stands today, bots are nascent. As with any technology, bots will need to mature before we can gain a clearer view of the role they’ll ultimately play and their lasting impact on both our everyday lives and as a tool for business. Most bots in use today aren’t demonstrating any meaningful form of AI. They are instead carrying out a very narrow range of specialized, niche tasks — ordering pizza, checking the status on a shipment of new shoes, or determining whether a flight is delayed. It’s worth noting that within this narrow scope the bots tend to get the job done, but their potential is so much greater. So the question becomes: What will it take for bots to take the next step in their maturity process and reach that full potential?
As Apple, Microsoft, and others have learned the hard way, one of the major challenges in building an all-knowing, monolithic bot that can handle a wide range of tasks is that it’s massively complex. This shouldn’t come as a surprise. The ultimate goal of these bots is to interact with us as naturally and effortlessly as a human would, but that means they need a deep understanding of human language and interactions — one of the most daunting tasks we can ask of a machine. And if we ever can get there, it would take longer than today’s businesses are willing to wait. What we need, then, is a different approach that can help make bots perform better and faster. And once we find that solution, we need a plan to successfully integrate bots into different parts of the enterprise, including — and especially — the support team.
Enter the bot colony
The technical solution: bot colonies. If bots are expected to interact like humans, doesn’t it make sense to organize them in an occupational structure similar to how we (and bees) have worked throughout the ages — by assigning specific roles to groups of bots, in the name of having those groups work together as a collective team?
We’d start with Worker Bots, whose specific focus would be on individual, rote tasks that can be executed quickly and simply. They would have a simple and well-structured language that they use to communicate. Within this group there could be the Returns Bot, the Order Bot, the FAQ Bot, and so on. These bots all would have very limited syntax and use card deck-style or quick response messages to solicit input. The scope of responses they deliver is deliberately narrow so that there’s no risk of them getting confused or wandering outside of their knowledge domains.
Overseeing the Worker Bots would be the Queen Bots, the controller bots that direct traffic to other bots and issue commands. The Queen Bots’ primary function is to interpret every customer query so they’re taking the higher order language and figuring out which other bots to bring into play. For example, a Queen Bot wouldn’t know how to process a “return” query, but can determine when the conversation is about “returns” and can direct the customer to the Returns Bot.
Finally, at the top of the chain is a human, who can step in and handle any queries that fall beyond the scope of ability (or knowledge domain) of the Worker Bots or Queen Bots. In many systems bots can appropriately handle a large percentage of customer queries; for example, in the customer support world this can be at least 80 percent, leaving the remaining and more complex 20 percent of queries to humans. By assigning the work this way, there is a clear chain of command and every member of the bot colony is playing to their strengths and skill sets, all without the customer seeing the machinations taking place behind the scenes.
This tiered approach to bot deployment not only streamlines the customer query response process — thereby delivering optimal customer service — but also serves as a valuable blueprint for businesses that want to implement bots quickly and add functionality as needed. With this approach, adding functionality is as simple as incorporating a new tier or type of bot, once it’s equipped with the appropriate level of AI for its task, to deliver a top-notch customer experience.
Putting bots to work for business
The second challenge of utilizing chatbot technology is find the best way to put these bots to work. Simply put, businesses that think they can replace customer support agents with bots are shortsighted. And in fact, implementing chatbots too early can actually damage customer satisfaction. The best way to implement bots quickly is to do so with the expectation of making existing customer support agents more efficient.
Picture a customer support agent who’s managing live chat for an e-commerce business. This agent is getting a variety of questions about products, shipping times, status of orders, etc. A bot can support this agent by reading the chat content quickly and suggesting a likely response while also suggesting the likelihood that this will be the appropriate response. For example, if a customer is asking if an order has shipped, the bot should be able to recognize that as a common request, scan for an order number in the chat and then check it. Perhaps the likelihood that this response is correct is 90 percent. But in another case, a customer might be giving an order number and asking if the package has shipped because they want to return it. In that case, leaving the response to a bot may miss the point entirely. While the chatbot learns through trial and error, governed by the customer service agent, the tool can help the agent respond more quickly to common requests. In most cases, this follows the 80/20 rule — the bot can help with 80 percent of requests that are common, while the agent can focus on the 20 percent of more complex questions or issues.
The answer to unlocking the potential of bots interacting with humans, as well as acting just like humans? Have them work like humans: as a team, with each one playing to their strengths.
About Agent.ai
Agent.ai provides artificial intelligence-powered CRM solutions that help businesses connect in real-time with their customers wherever they are, whenever they’d like and in the communication channels they prefer. Backed by its event-based machine learning and natural language processing engines, Agent.ai enables businesses to respond to customers faster, 24/7/365, while helping agents become more productive. Visit www.Agent.ai to learn more.
