Chatbots are making a comeback. They have been a hot topic over the past year or two thanks to technological breakthroughs that are making them smarter and faster. In fact, chatbot technology is so hot that Gartner believes that by 2020 you will be more likely to chat with a bot than with your spouse.
The idea of a smart, helpful chatbot being the modern-day personal assistant is enticing to most workers and businesses. But, before we get that personal with bots, they still have a long way to go.
Where chatbots need to step up
In order to fit the bill as a true productivity booster, chatbots need to advance their natural language processing (NLP) and artificial intelligence (AI) capabilities. With these enhancements, bots will be able to better understand the intent of a conversation and interpret humans’ emotions during it. Intent is the missing puzzle piece that, once they can grasp it, will give bots the context they need to provide personalised, relevant materials when we need it, making us more productive.
NLP still has a few challenges when it comes to turning complex requests into commands a chatbot can understand. Because chatbots can’t comprehend the nuances of intent or emotion from hearing human speech, they are not able to grasp the full meaning of many phrases or commands in a human capacity. This presents an issue for customer service chatbots in particular – if they cannot interpret these important context clues and, for example, know to route an angry-sounding customer to a live representative, then putting your customers in the hands of a chatbot could result in frustrating experiences.
Interoperability will soon become an issue as more new chat platforms are becoming available. Because there have not been many standards or guidelines created regarding how chatbots interact between various chat platforms, there is a very low probably that different platforms can understand each other. As such, chatbots will have to be re-written for the platform in which they’re deployed. With different departments of large enterprises using different bots and chat platforms, migrating to one platform that will act as single source of truth will be challenging to implement.
Today’s bots are extremely limited in their ability to respond with dynamic content. That’s because they are not programmed to respond with a secure integration to a transactional business application to provide a rich, helpful answer. Typically, chatbots only know the correct responses to a narrow set of questions.
The level of knowledge bots currently possess is very basic and narrowly defined – think of teaching a dog how to fetch your slippers. The animal can easily learn to repeat a basic and narrowly defined command. They know how to identify exactly what your slippers look like and where you usually keep them. But the moment you want to do something more complicated, like ask for your running shoes from the back porch, there will be some confusion. The same is true for these basic bots – they can reliably do simple tasks you teach them, but anything more sophisticated will require enhanced development.
As AI becomes more advanced, it will help chatbots better understand intent. With this knowledge, the bot could brainstorm new solutions that might not have been designed by the developer. For example, if an employee was using a chatbot to book a conference room for a team of 10, then a chatbot could find rooms with 10 chairs during the selected time. An AI-enhanced chatbot could learn from prior interactions that there are often last-minute guests to meetings and encourage the employee to book at room with 12 chairs to fit everyone comfortably.
Current use cases for chatbots
While chatbots have a long way to go, there are three business scenarios that can benefit from chatbots in their current state.
On-demand Customer Interactions
For some businesses, it makes sense to have a chatbot on the front-line of support or Q&A help desk to provide a customer with an immediate response. We’ve seen many customers starting with simple chatbots that use a search index to match a user’s question about a product or service with a known set of knowledge-based articles. The key reason to use chatbots for this function is that they never get tired – they are available 24/7, making them an ideal way to reduce support costs while improving customer satisfaction. However, it’s smart to have a safety net of human customer service agents on-hand for the bot to pass off a customer if they can’t understand their needs.
Personalised Notifications and Alerts
The parameters of notifications and alerts will change employee to employee based on their role, responsibilities and department. Chatbots can be very useful here because they can take a set of parameters in plain language and automate an alert back to the employee. Such alerts can be taken a step further by appearing on an employee’s mobile device, making it easier for them to pay attention to important alerts, respond to them and increase the ROI of existing transactional applications.
Automating Data Entry & Search Retrieval
Not all business processes can be easily replicated by a chatbot, but good candidates to be automated are those that require minimal fields to create a record. Instances would include filing an expense report receipt, uploading an image of a business card, or registering an email address to a mailing list. In the same vein, the search and retrieval of specific records can be streamlined with a chatbot. For example, a sales rep could ask a chatbot to retrieve the address of the next customer she or he needs to visit. Automating these processes via chatbots enables employees to spend less time on administrative tasks and more time on the strategic and creative aspects of their job. It also increases the rate of adoption for existing transaction systems where these processes and data are stored.
How to implement chatbots
Now that we have the lay of land on the state of chatbots and what they can be used for, how does a business go about implementing these sci-fi algorithms into daily practice?
For companies just now exploring how to integrate bots, they should start small and set reasonable goals for using the technology. It’s advisable to identify a function that is not yet business critical – where you can afford to take the risk of bots giving an inaccurate answer. Once the process or department is identified, test and measure with lots of qualitative and quantitative data. The initial findings from these tests can be used to continue broadening new chatbot use cases.
It’s also smart to give chatbots transactions that are typically handled by other bots, creating bot-to-bot scenarios. In this case, be sure that this is a low-risk task as it’s likely that the bot’s understanding of language is limited.
As with most fads, most companies have yet to determine if chatbots have provided lasting value. NLP and AI mark our entrance into the possibility of a truly smart chatbot that can add business value, rather than causing confusion and distraction.
Stephen Hamrick, VP Product Management, Collaboration Software, SAP
Image source: Shutterstock/Montri Nipitvittaya