The Time to Reinvent Chatbots
Chatbots have been hot. In the tech world, they have been spicier than a scotch bonnet caramelized in vindaloo and dipped in jalapeño sauce. But the sizzle is now going out.
Chatbots facilitate natural conversations without the expense of human capital. Whether for brand-building, customer acquisition, product discovery or support, chatbots offer awesome potential, infusing empathy and intelligence into the core design of every brand-customer interaction. It should be no surprise Facebook Messenger already hosts a chatbot community 35,000-strong.
Marketing managers aren’t the only ones who should be excited about the promise of the AI-powered chatbot. From delivering your pizza to reminding your kids to do their homework, the tasks chatbots are capable of automating are getting wider all the time. Most excitingly, chatbot functionality can be rewired to work internally – to oil the gears of an enterprise so it becomes leaner, more agile and smarter than its competition.
The current generation of chatbots are not designed right. But don’t let that put you off. This article talks about how to make useful chatbots using cognitive computing.
Think of a good enterprise chatbot as a supremely-talented subject matter expert, and it’s subject matter is your business. It can be deployed anywhere in your organization, at any time, so any worker can interact with it to get the information they need, or suspect will be useful. Applications of chatbots range widely:
- Automating knowledge discovery and data mining. Chatbots provide an intuitive and intelligent means with which humans can naturally interact with an organization’s data. If you want to capitalize on the mountains of documents and big data piling up inside your enterprise, chatbots are much more effective in implementing natural language search than traditional search interfaces
- Automating complex standard operating procedures, like the onboarding of new staff. A chatbot can be trained to know every single line of all of your enterprise SOPs, and be able to recall and provide that information to workers in an instant. This frees up key decision-makers across the business to spend less time hand-holding junior recruits and more time using their expertise, creativity and judgment to drive real business value.
- Automating tasks using a friendly, forgiving interface for customers. Chatbots, if designed well, can really be a completely new interface for the enterprise with its users. Such chatbots can unburden the enterprise with a lot of such tasks, like answering questions, filling up forms in an interactive way, or understanding getting simple things done – for example travel flags for credit card companies.
But 70% of requests made to chatbots on Facebook’s Messenger platform fail – and, for an enterprise, those numbers simply don’t work. No business manager in their right mind would entrust mission-critical roles or tasks to a worker making fundamental mistakes so consistently.
That being said, you shouldn’t let the badly-designed bots of the consumer fad put you off – chatbots can make sense for real business too. We just need to build them in the right way.
Cutting through the chatbot hype
The reason most chatbots fail is most chatbots can’t really chat, or interact, like humans do. Based on keyword correlation, they are unable to understand the real relationships between words, or the context behind them.
Humans are able to chat so easily because we process things as ideas, rather than keywords. We know, for instance, that ‘black box’, ‘data log’ and ‘transparency’ are related ideas. If you ask a typical chatbot to identify all of the black boxes in your organization however, you’ll see that it doesn’t. It might understand ‘black box’ as business jargon for an opaque device or system, but it would not be able to intuitively turn that information into something valuable, like a transparency score or data log. Because a chatbot can’t understand those inter-relations, it can’t retrieve the information a human really wants, or perform a task a human might.
So, how do we build a chatbot that is more than a keyword monkey? How do we build something that really reduces friction; that helps workers achieve specific goals without needing another human to be involved?
How to build an enterprise chatbot
To get a chatbot to effortlessly inter-relate ideas in a way that streamlines tasks – i.e. makes them easier and quicker to complete – we need it to emulate human thought process. By thinking like your high performing human capital, it can efficiently automate some of their tasks. Thinking like a human means:
- Focusing on ideas, rather than words: In human conversation it is always important to realize the similarities between various ideas like “a red ball”, or “a maroon sphere”, or “the planet Mars”. A chatbot must also process ideas without the underlying keywords.
- Prioritizing ideas: Not every idea in a context or a document is equally important. Well designed chatbots need to differentiate between the ideas for credibility and significance. For example, “Telephone” is the central idea of a wikipedia article of that title, and is almost meaningless in a call log.
- Learning to say No: The most effective workers aren’t afraid to ask for help or admit knowledge gaps. Chatbots that straight away say “I don’t know”, give the organization maximum time to find an alternative solution. We believe learning to say no is the most fundamental building block of a high reliability chatbot.
Using these three design imperatives, it is possible to deliver chatbots for various applications and Fortune 100 companies with 98% accuracy. The bot can interpret ideas without relying on keywords. Order ideas by importance, then store them into a hierarchical data structure. And even build contexts from live conversations. The best answer to a human question like, “who in legal is responsible for approving this task?” is built using the context and accessing the hierarchical store of ideas.
These consumer and enterprise solutions offer a huge jump in capability and performance from most Facebook Messenger bots, and it’s all down to one branch of AI technology: cognitive computing.
The “cognitive” part is how our tech makes sense of inputs traditional computing approaches have a hard time with. Utilizing a combination of natural language processing, real-time computing and machine learning algorithms, Coseer chatbots can decode the messy, ambiguous outside world, and structure the resulting information so humans – or more traditional computer programs – can take action.
The future of the chatbot
Although a good start, chatbots built on the keywords design framework were nowhere near enterprise grade. What just about worked in the customer service world was nowhere near good enough in the cut and thrust of business.
But now AI is integrating ever more deeply into our lives – personal and professional – idea-oriented chatbots can be a key tool for assimilating the human and computing worlds. Coseer’s solutions show we are already capable of designing and training machines to process information like humans do, talk like humans do, and provide business value like humans do.
Since these bots run round the clock, at a fraction of the cost of a human resource, without any manual or fatigue-related errors, cognitive bias or risk, it’s no overstatement to say that the future of the chatbot could be the future of all business.
Want to learn more about how chatbots can integrate and automate in enterprise? Book your Chatbot for Enterprise Workshop here.