The 200 billion dollar chatbot disruption

In 2014, Facebook acquired WhatsApp for $19 billion. That astronomical number set off waves of speculation as to what value Facebook could possibly see in a company with just 55 employees and roughly $20 million in revenue, although it had 500 million users. At last week’s F8 conference, that vision became a lot clearer, and it’s big. Chatbots will cause a near-term disruption in how businesses interact with consumers, and a long term paradigm shift in how people will interact with machines.

Consumers and chatbots

The easiest way to see why chatbots will make a near term impact on everyday consumers is by comparing a modern day customer support call to a chatbot experience. Today’s real time customer support options are typically either through voice, web-based chat, or Google search. To illustrate the inefficiencies here, consider the following set of steps and substeps in a “1-800” voice based customer support interaction:

[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":1938176,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,commerce,mobile,","session":"D"}']

1) Find contact info of company through the Internet
2) Call contact number
3) Navigate through a sequence of touch tones or semi-functional speech recognition steps
4) Wait on hold until a specialist is available
5) Step through a sequence of verifications to make sure it is truly you who is calling
6) Get told you need to talk with a different representative and get put on hold while you are transferred
7) Connect with a rep who now tries to have you navigate the Internet via voice instructions
8) Login to the company website
9) Finally, get an answer to your question

Here’s how it works with the chatbot experience:
1) Open Facebook Messenger and search for the business name using its “handle”
2) Start the conversation by making a request
3) Receive rich media feedback (text + images + hyperlinks + voice) that answer your question

To highlight the key differences, let’s itemize the various gaps in the two approaches:

• Chatbots offer a standardized way of connecting with businesses (similar to Twitter handles or email addresses), reducing the complexity of finding the contact info for a company.
• By integrating with Facebook identity, chatbots allow for automatic verification of personal information (no more cumbersome account lookups or credit card entries).
• There is no experience of “transferring” in the chatbot scenario; if another expert needs to connect, it can happen seamlessly in the background.
• In the modern world, the majority of information dialogue revolves around the Internet. Chatbots allow for information to be transferred via hyperlinks and rich multimedia rather than through a linear voice translation.

In the other scenarios of web-based chat or Google search, there are similar friction points that chatbots overcome through the same principles as illustrated above. Simply put, chatbots offer a superior customer experience, and businesses will quickly realize that if they don’t offer chatbots, their competitors will, and they will be at a serious disadvantage. But that’s not the only reason companies will invest in chatbots.

Businesses and chatbots

Beyond offering a better customer support experience, businesses will soon find that chatbots allow for dramatic cost savings in their call centers.

The main driver behind the cost reductions here will be through advances in natural language processing (NLP), supplemented with crowdsourcing approaches. Let’s start with NLP, which is a form of machine learning. The simplest way to think about NLP is as an automated way of understanding groupings of words and the corresponding natural responses to those groupings of words. For years, the challenge with NLP applications in the business world has been that companies have tried to solve the automatic speech recognition (ASR) problem first or in tandem with NLP, which is notoriously difficult.

The brilliance of the chatbot movement, is that it decouples NLP from ASR and allows for significant business value to be obtained without needing the ASR piece of the equation. NLP is still a hard problem to solve, but it’s solvable to sufficient accuracy with modern day machine learning approaches. The problem becomes more tractable when you consider the approach of augmenting NLP with crowdsourcing. Crowdsourcing allows for simple tasks to be divvied out to a distributed workforce. These “microtasks” can be performed by anyone, whether that be an expert customer support rep within the company or an anonymous worker in the cloud.

Combining the two techniques, we see that chatbots will allow portions of a customer support conversation to be handled via NLP, and when there is a statistically significant indicator of incorrectness, a human can be looped into the conversation to make sure edge cases are handled. And this can be done seamlessly, without the end user even realizing it.

[aditude-amp id="medium1" targeting='{"env":"staging","page_type":"article","post_id":1938176,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,commerce,mobile,","session":"D"}']

All of this leads to a virtuous cycle called a “closed loop” in machine learning systems, where the human answers help train more sophisticated algorithms as more business use cases are provided. The result is a spiral effect where the NLP accuracy continues to increase, which in turn continues to lower human costs as more sophisticated algorithms are able to answer harder problems.

Now, let’s put this together. NLP allows for automated customer service, and crowdsourcing lets you maintain quality by making sure there is always a human present when needed. A good rule of thumb with modern day machine learning approaches is that they typically can provide accurate responses that gradually taper off at 85% as more data is provided. This sets a significant theoretical cost saving that companies can gain by leveraging chatbots+NLP+crowdsourcing. (Theoretical because there are other considerations when using machine learning techniques, to be addressed in future posts).

Next, companies will quickly see that by embracing chatbots for customer support, and driving down their “cost centers”, they will have opened up a direct dialog with their consumer, which will pave the way for marketing and sales to open new revenue channels. Since the customer support interactions with chatbots are no longer through an anonymous 1-800 number or web chat, companies will know who their customers are, what their customers want, but more importantly how to directly engage their customer. It then becomes a simple cost optimization problem to close the sale. Again, much of this process can be automated with NLP, assisted by human sales reps.

After the sale, the business can continue to engage with its customers through marketing campaigns aimed directly at the customer’s needs (and there goes Twitter’s market cap). At this point, the interactions between businesses and consumers will have fundamentally changed, and so will the way in which people think about how they interact with machines.

[aditude-amp id="medium2" targeting='{"env":"staging","page_type":"article","post_id":1938176,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,commerce,mobile,","session":"D"}']

Chatbots and the future of human-computer interaction

This is where things get interesting. In the long term (5+ year outlook), all of this will lead to a treasure trove of data for the messaging platforms that facilitate chat interactions. Facebook and WhatsApp will be able to use this data to develop better tools and rich datasets to bring about more advanced NLP engines. Think of this as being analogous to what Google has done with search; the end users don’t even think about it, but the datasets that Google has collected are what really make search light years better than when search engines first surfaced.

This is where we’ll see personalized AI really take off, and where things like Internet search can fundamentally change. When NLP technology advances to the point where the need for human oversight is minimal, we’ll see personal assistants that use the Internet as a data repository from which they can retrieve pieces of knowledge that more cerebral humans can put together for higher value decision making. The reason humans use search engines today is because they have no better options, not because there is an inherent value in spending time sifting through search engine results. NLP moves us one layer up in the information gathering value chain. When this happens, Messenger and WhatsApp will usurp Google to become the entry points to the Internet.

Implications

Chatbots are poised to fundamentally change the way humans interact with machines within a five-year horizon. The starting point will not be personal assistants, or chatbot “apps”, but rather the investment that businesses make in order to drive their topline and bottomline metrics (starting with support, and moving into sales/marketing). This will lead to a treasure trove of data that will allow for further disruptions in how humans and machines interact and will completely change the way people interact with the Internet as we know it today.

As an emerging technology strategist, I am fascinated by many technology advances coming from Silicon Valley. Self driving cars, augmented reality, and the blockchain all have major disruptive implications over the next 5-10 years. The reason I have put chatbots at the top of the list is that I believe we’ll see major traction not five years from now, but by year’s end. It truly feels like we are in 2008 and the iPhone app store has just been announced (in terms of magnitude of disruption, not productization). What will follow with chatbots will move so fast that people won’t even realize what’s happening until it’s right in front of their faces.

[aditude-amp id="medium3" targeting='{"env":"staging","page_type":"article","post_id":1938176,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"bots,commerce,mobile,","session":"D"}']

Matt Swanson is Managing Partner of Silicon Valley Software Group. He advises a number of companies on their technology strategy and serves as the acting CTO of others. He also works with enterprises to help incorporate innovation into their organizations. He previously founded SpeakerText, which raised from Google Ventures and 500 Startups before being acquired in 2012.