Intelligent Messaging Is Getting Smarter, Everyday

Estimated reading time: 4 minute(s)

Chatbots can be a smart way to interact and engage with your customers especially through messaging apps, but you’ll have to train them.

Since the beginning of 2015, there have been more monthly active users in messaging applications than in social networks. This fact, along with the advancements in machine learning (ML), artificial intelligence (AI) and natural language (NLU/NLP) technology, has created a sustainable business case for leveraging automation (chatbots) in all messaging channels.

Smarter Messaging

It’s no secret that device and platform companies have been making messaging channels smarter for years. Now, messaging channels offer numerous features including location , auto­correct (albeit, at times in funny or more annoying ways), automatic emoji options & autofill, and commerce. The smartest addition by far though, the one that will revolutionize how companies are using messaging with their customers, is integrating chatbots into the messaging customer experience and leveraging AI & machine learning.

Chatbots are poised to radically alter and improve customer interactions within messaging apps.

Chatbots have been around for a while. The term chatbot dates back to the late 60’s early 70’s with chatbots ELIZA and PARRY that used keywords to provide specific answers. These keywords are called “intents” and are still used today to make chatbots smarter by teaching them how users request information.

Self­-starting Smarts?

But the real question is, can they get smarter––on their own?
The short answer is yes, but one of the biggest misconceptions is that a chatbot can be smart enough to handle unlimited queries from day one. To understand this, think of a chatbot more like a new employee.

Let’s say you have a CRM company and you hire a new sales rep with previous CRM software sales experience. While they may have been number one in sales at their previous company, is it realistic they’ll know everything about your product to begin selling their first day on the job? (Obviously not.)

Just Like a Human Employee, Data Training is Imperative

I often advise companies to orient to chatbot technology the same way you would when on­boarding a new employee. There should be training, education and an appropriate learning curve that allows for missteps. Over time, the employee becomes more savvy and begins to do their job more effectively. They gain valuable on ­the ­job training about actual processes and how their customers use their product. Chatbots require similar training, or “ data training .”

Creating an intelligent chatbot takes time, similar to the neural network formed over time from childhood into adulthood, in which the most useful pathways generated through experience are tested and retested. Yes, there may be thousands of well­-programmed intents (again, what a user types into a messaging application to interact with a chatbot), but these then need to be integrated into the chatbot’s application lexicon to make it smarter.

To really guide your customers on a delightful brand journey, your chatbot must be programmed to take customer signals and feedback and make these useful for your customer ­­just as an experienced human sales executive does, that is, it needs a neural network and some training.

The Smarter the Chatbot, the More the Data Training

We would all agree that IBM’s Watson is one of the elder statesmen in the AI evolution. But how did it get there? When they first developed IBM Watson, its “brain” was empty. But today, Watson has a deep knowledge base.

To train Watson, IBM programmers used encyclopedias , dictionaries, thesauri, newswire articles, and literary works from numerous databases, taxonomies, and ontologies — specifically, DBPedia, WordNet, and Yago. The IBM team provided Watson with millions of documents and other reference material that it could use to build its knowledge. To process and access this data efficiently, Watson needs a lot of power, memory and energy to operate – A cluster of ninety IBM Power 750 servers, each of which uses a 3.5 GHz POWER7 eight­ core processor, with four threads per core. In total, the system has 2,880 POWER7 processor threads and 16 terabytes of RAM.

In other words, it is the Mother of All Chatbot Brains.

Jeopardy IBM Watson
IBM Watson proves you can train for Jeopardy.

The Smartest Chatbots are Leveraging Neural Networks

Most people who aren’t experts don’t really understand neural networks completely (myself included) but simply put, they use two methods, data which is feed into them and data which is generated by learning independently.

In the first case, a neural network is fed data without any actual experience necessary. In the case of the AI programmed to master the game GO, for example, the neural network was fed as many as 30 million moves from games played by human experts. In the second case, neural networks can also be used to make the brain smarter experientially. The same GO AI enabled the computer to play thousands of games with itself to learn from that experience.

Data needs to come from both pre-­existing and new experiences for chatbots to truly learn. Existing data first needs to be catalogued and integrated into the brain. This will define the structure of the chatbot. Then, by using new experiences, the chatbot can begin to learn from them while synthesizing existing data to make itself smarter.

For any company thinking about integrating a chatbot and AI into their customer experience, they should understand the level of work involved, but also that the opportunity of leveraging AI within these customer conversations will be revolutionary, not only as a practical resource improving their customer experience, but also as an employee, learning and improving with every interaction with each and every customer.

Previous articleCortex Innovation Community Celebrate Aloft Hotel Groundbreaking
Next articleHive Network Offers Startups A Clear Path To Developing Their MVP
blank
A pioneer of the early days of messaging, Michael combines the latest trends in technology to create products that transform the messaging space and help to define its optimal direction. Michael is the co-founder and CEO of nativeMSG, which leverages AI, NLP and Machine Learning to create intelligent messaging strategies integrated across multiple messaging and voice platforms.