IBM has recently estimated that customer service chatbots will help businesses all over the world save $8 billion per year by 2022. As 2019 has just begun, more organizations are expected to make efforts and investments into adding AI-based chatbot and intelligent assistant solutions to their products and service portfolio. While companies have yet to realize the true business value from leveraging artificial intelligence and machine/deep learning tech, they’re already facing certain challenges.
Forrester points to the following Top 3 challenges of AI adoption and value generation:
1. Data quality
“Regardless of your beginner or expert AI status, data is the Debbie Downer of any AI project. While enterprise aspirations for AI run high, in 2019, we predict continued investments in good ol’ information architecture (IA). The tables will turn from AI to IA in the majority of firms that have already dabbled in some form of AI, as they’ll quickly realize that their irrational exuberance for AI adoption must be equally met with solid efforts on an AI-worthy data environment.”
2. AI and data talent shortage
Two third of IT decision-makers struggle to find and attract qualified AI and Big Data talent, while 83% struggle to retain them.
Yet, talent shortage goes beyond data science and technical skills; today’s organizations need to focus on finding specialists capable of creative problem solving and resilience over coding and statistical skills.
“Traditional recruiting practices fail, causing companies to seek out new approaches and tools. In 2019, we’ll see firms start to tackle the AI shortage by applying AI to recruitment.”
3. A happy marriage between RPA and AI
The robotic process automation (RPA) gained momentum before AI caught on and piqued the interest of enterprises. Forrester predicts that AI and RPA will join forces and turbocharge the innovative efforts of companies.
“Firms are already combining AI building block technologies such as ML and text analytics with RPA features to drive greater value for digital workers in four use cases: analytics that solves nagging platform issues; chatbots that boss around RPA bots; IoT events that trigger digital workers; and text analytics that lifts RPA’s value.”
What AI decision-makers think about the business value of chatbots
Attempting to evaluate what IT decision-makers think about the business value of AI-driven chatbots, I reached out to them on Quora and got some thought-provoking answers that are worthwhile to be shared here.
Alan Tan, Sr. Director of Machine Learning/IoT/BlockChain at SAP, says that chatbots have been around way earlier than the term was coined.
“The very first known chatbot Eliza actually revealed a lot about both ends of the “chatting” and provided lots of insight into human-computer interaction.
The key is human will interpret what the chatbot says on their own experiences, and thoughts, way beyond what technically the bot is trying to express. This means humans can interact with chatbots at a level above what chatbot actually “understand”. This part of the interaction is a feature of human, and it was demonstrated way before AI was part of chatbot – Eliza was a (fairly rudimentary in today’s standard) rule-based text manipulation program that takes turns to accept input and prints output.
Understand this point is very important for using a chatbot in any environment especially in business.
The main difference between AI based chatbots and rule-based chatbots is the ability to process language components – extract intents, entities, establish and use contexts, etc. (Chatbot is a very notable use case of NLP). So AI-based chatbot can manage lots of information processing exchanges quite similar to how human exchange ideas now.
In terms of how business benefits from AI-based chatbots, first, at the current technology level, a blended AI + human chat system can actually handle most business chat seamlessly while a majority of the customers do not realize which part is handled by AI and which part is by human – typical customer service use case show 90% or more requests can be handled by AI portion, and actually customer satisfaction rating is higher for the majority of the requests compare to human agent (although this is not a fair comparison, because human agents handle more challenging issues).
The multi-channel chatbot for customer service is a fairly robust field now – as people switch to electronic endpoints from the traditional analog voice call, the wait time for the customer (the number one complaint about call-centers) dropped significantly, and consistency among channels (what people hear from one call-center agent and what they got in an email from another agent) is maximized.
The chatbot can replace (or at least supplement) most of the support/request ticketing systems.
I actually cannot imagine any business with a significant number of customers does not use a chatbot in some sort to provide customer service these days.
Another “low hanging fruit” for a chatbot is knowledge management – while search had been the previous revolution of knowledge management, searching is actually getting more and more language like. And multi-round natural language inquiries are much more natural for humans than structured library indexes. There are use cases where companies move their internal policies into the chatbot (not a small project though), which makes a big difference – instead of having new employees read and interpret all policies and come up with their individual judgment on how to deal with certain situations, chatbot based policies can help “answer” questions regarding how to treat certain situations based on the chat the employees asked.
A little bit more sophisticated than “use chatbot”, some applications are also moving toward a “conversational UX”, instead of using GUI, or reports, applications that used to provide a concise output, are benefiting from chatbot-ish UI, basically, reduce end-user training to ~0. Think of what Alexa does at home, in a business environment, we can do quite similar things – i.e. placing an order, getting approval, tracking the shipment, can be done in a chatbot, or an Executive reading financial report can use conversational UX to drill down or aggregate metrics and KPIs.”
Eddie Baltrusaitis, Chatbot Product Manager at ChatStud.io, has his take on the subject matter:
“AI-based chatbots are used by business giants that need incredibly wide variations of a technical solution.
These type of chatbot converts information received from a user into an understandable format and store it in a knowledge database by learning and leveraging preexisting knowledge and acquires it continuously. Based on this decision, the chatbot takes action to achieve predefined goals.
So in general, very few businesses need and can afford this type of bots, most of the entrepreneurs use rule-based chatbot model that works perfect (you can literally adapt this type for every business niche on the market) and in way cheaper way.”
To learn more about how AI-based chatbots help drive enterprise value, please check out my previous article on Hacker Noon.
Chatbots, both AI-based and rule-based, pop up like dandelions, with the overall customer reactions being pretty positive, but still somewhat apprehensive. One recent study has found that although chatbots are believed to be faster than apps, still they aren’t as convenient.
As businesses learn more lessons from using chatbots, they’ll have a better understanding of how to generate more value from them.
And do you have any firsthand experience building and using AI-based chatbots? If yes, how did it improve your top-line growth or bottom-line savings? Please share your success or failure story in the comments below.
