In researching this month’s topic, I tripped across this hysterical movie match up contest set up by movie reviewer Aaron Peck back in 2011. The contest asked readers to vote on who would win a battle between swordsman Inigo Montoya going up against the sentient machine HAL 9000 – asking the question “Will Inigo’s left handed swordplay be able to withstand HAL’s all-seeing eye?” (see contest here)
Having designed and run multiple global customer support organizations in my career, I’ve always been obsessed with new and interesting ways to provide customer service through online channels like Chat. Now I see a massive up swell in companies seeking Digital Transformation, and the recruitment of AI Bots to do the ‘dirty work’ of customer support. This prompts me to spend some time exploring the future CX and the world of digitally transforming businesses as they compete against digitally native businesses that were built for disruption.
For me, “Digital Transformation” basically means taking any company founded before 2004 and transforming it into a digitally savvy, customer-centric machine that consumes mass quantities of interaction data to more efficiently deliver the value customer’s seek. The process of digital transformation is crucial to any company that wants to thrive in 2017 and beyond. As I’ve said before, CX obsessed companies know they are racing UP the DOWN escalator due to the flood of great digital CX happening every day. Forbes recently stated that 84% of digital transformation fails hurt and occasionally critical wound the business. It is a journey that will challenge and test every element of their existing business model.
I’ve been thinking about digital transformation a lot, and in particular, the use of machine learning based artificial intelligence bots. Bob Thompson, the editor-in-chief of CustomerThink, recently wrote, “It is predicted that by the year 2025, up to 95% of all customer transactions will be AI-powered”. As we know, good CX is based on tangible and intangible exchanges between a customer and company that deliver on a brand promise. So how do we as customer experience professionals train these bots to have authentic, empathetic responses that resonate with our brand promises?
This is where it can get complicated. Machine Learning, Artificial Intelligence, etc. are all specialized methods, which generate computer code and mathematical functions (models) from large sets of data. The process of generating models from historical data is called training. Using different algorithms and methodologies, models can be generated to classify, predict, recommend, or optimize future interactions. The best practice in machine learning is to train the algorithm/models by analyzing massive volumes of customer touch point and transactional data records captured over a number of years. Michael WU Ph.D. Chef Scientist at Lithium Technologies tells us that “one of the most common arguments favoring big data is that data is versatile and doesn’t have a shelf life. Even though you don’t need it today, its relevance and utility may become apparent in the future.”
The capture of data and the ability to use it to drive your business forward is key to digital transformation success. If you read my previous blogs, you know I often coached leaders on the fact that their “companies actions betray their intentions.” You can have the best brand marketing in the world, but at the end of the day, the way you treat customers and the front-line associates who service them will result in two ways: Long-term business successes through loyal and engaged customers, or short-term business failings through high customer churn resulting from gimmicky customer acquisition.
One of my current projects is developing a digital curated experience for customers to access services. Part of that development includes sifting through years of historical data and feeding that into a data lake to create the business rules that trigger good or bad AI CX interaction responses. I need to consider the fundamental details when teaching a learning machine. A human CX practitioner would probably understand that asking a customer for an up-sell/cross-sell offer right after ordering their 1st product or mid-stream of a customer service interaction, is a bad idea. If this type of disconnected marketing offer happened frequently over the past couple of years, this bad behavior is embedded in the data upon which the AI acts. These bad practices and potentially ‘off brand’ calls to action will be utilized, and the negative bias is now institutionalized and executed with digital speed and precision. So how do I teach a machine to overcome bad behavioral bias and deliver better CX?
Looking at success rates in the customer lifecycle is crucial when figuring out what to teach the machine. Both the positive and negative customer/business outcomes need to be considered, and customer cohorts should be formed when selecting the right historical data to generate the machine learning. We have to find the patterns that promote the deepest customer engagement over time. The big idea around AI Bots is that they can anticipate the best thing to do when interacting with a customer based on the analysis of massive data sets. What I have learned first hand are these three simple guidelines:
1) Feed the machine good CX outcome data to create positive results, as well as teach it what not to do based on the bad CX outcome data.
2) Focus the rules of AI interaction towards successfully navigating the five micro value gates (See my March 2017 blog), not just the next sales transaction.
3) Use consumer validation testing and live interaction tracking to understand when to switch over to a human interaction (off ramp) if the AI bot goes off script. The machine learning algorithms can help you know the next best action success predictability tracking and based on a numerical threshold ( e.g. When less than 81% chance of success . shift to human chat agent)
4) Take a “Fail Fast Forward” approach that eagerly seeks to make NEW MISTAKES, to quickly learn from, and lean forward from there. This will help you focus on Build / Measure / Learn / (Repeat or Pivot) the AI Bot interaction experience to drive a set of specific customer success outcomes. If you can't measure the success of doing something, you probably shouldn't do it - so be clear about your CX Design goals and customer success thresholds that meets your operational and brand promises.
As a company gets serious about moving forward with their digital transformation, they need to pause to map and understand the location of their customer’s journey pain points. They must find moments of truth by collecting as much data as possible so the machines can understand which approaches lead to the longest customer satisfaction and look out for unintended consequences. For successful AI digital interactions, it is crucial to understand how the machine bots learn and can be trained over time. I will continue exploring this topic over the next few months, digging into exactly how a company might enhance and improve their branded CX when navigating a digital transformation.