CX^AI: Executive Briefing on Designing the Intelligent Experiences of Tomorrow

Dan Ptak,

1.05.2018

In December, we hosted a panel of cross-vertical experts from Solstice, Narrative Science, and Dice.com to lead an executive briefing on the current state of AI and its impact on customer experience.

The group covered a lot of ground as we dined on lunch from the executive chef at the Soho House in the West Loop.

The Chicago Event Panel Included:

  •   Moderator: Jared Johnson, Principal Digital Strategist
  •   Panelist: Karl Hampson, Managing Director, AI, Solstice
  •   Panelist: Mauro Mujica-Parodi, VP Product Management, Narrative Science
  •   Panelist: Simon Hughes, Chief Data Scientist, Dice.com (DHI Group)

Below are six of the key takeaways from the interactive discussion.

1) Even when it comes to AI, business strategy should precede digital strategies.

There was broad alignment across the panel on one of the most discussed topics of the day: sound business fundamentals shouldn’t go out the window when it comes to this new, and yes, shiny, technology. Like any great digital initiative, business strategy should precede the technology strategy.

The current AI climate differs a bit from previous technological waves in a couple of fundamental ways: the hype has become more feverish than ever before and the technology more complicated. Both are leading teams to spend a disproportionate amount of time talking about potential solutions and not about business challenges or problems.

Panel members talked at length about the problems they have previously tackled through AI and machine learning and about some of the most pressing issues for 2018: greater personalization, reduction of customer churn, creation of a better onboarding process for customers or employees, elimination of non- or low–value-add activities in the enterprise, better enterprise search, possession of better information at the point of decision-making, and enhancement of the knowledge of frontline employees (customer service reps, financial advisors, etc.) to better serve customers and to make better use of their time.

A great quote from the end of this portion of the discussion came from Mauro at Narrative Science, “If the CEO walks out of a board meeting and says what our AI strategy is, just hit the ground.”

2) When it comes to defining AI, there is anything but alignment around a common way of thinking and talking about the technology.

Yes, there are plenty of textbook definitions of AI and associated technologies, such as cognitive learning, machine learning, and deep learning, but because of the relative nascency of the field, there is still broad misalignment in the market regarding the “right” way to talk and think about the technology and its applications.

Solstice groups the enterprise impact of AI into three key categories: intelligent interactions (e.g., search and conversational experiences), intelligent decisioning (e.g., delivering more predictive and insightful information to inform decision-making), and intelligent operations (e.g., digitizing and automating operations with the aim of creating more efficient operations and more compelling customer experiences).

The panel agreed that each organization should either adopt a preferred framework from a favorite analyst, technologist or vendor, or work to define their own in the coming year.

3) What’s the best way to build trust in AI systems? The same way humans do: mean what you say and say what you mean.

The panel had several different, but equally valid, takes on this question:

As Simon and his team of data scientists at Dice.com started leveraging machine learning to better match job candidates to employers and vice versa, they recommended a transparency-based approach, going out of their way to walk internal stakeholders through the model and the inputs to ensure a baseline level of comfort was broadly achieved. Early on, they regularly identified and discussed specific examples of where technology was outperforming human judgment or basic recommendations.

With consumers, it’s much harder to do this at scale. In this situation, Mauro and the team at Narrative Science recommend a performance-based approach. There are certain situations like algorithmic investing in which an application can track and report on the performance of passive investments for, say, the last 3 to 6 months. The technology can then offer up a plain and simple number which dictates what the individual’s portfolio would have been worth had he or she followed the AI’s advice. While not necessarily the most positive approach, this feeling of loss aversion is deeply rooted in our psyche and extremely effective.

Third, many machine learning applications rely on massive pools of data from large numbers of individuals. Overtime, these applications optimize their algorithms based on inputs from the proverbial crowd. It’s possible to create an environment that fosters a group mentality, wherein consumers derive their confidence and trust from the collective wisdom of the crowd. Amazon’s AI-driven product recommendations engine is a prime example of this approach. The thinking goes, “If I like a certain product, and everyone else who brought it also likes another product, then by logical extension, that other product must also be something that I’ll like/need.”

Finally, there was broad agreement that even today, most consumers care less about the idea of trust than they do about the overall better experience, especially when the transaction doesn't involve sensitive personal information. A more relevant ad, a more preferred style of jeans, a better neighbor recommendation, a better vendor recommendation, or more resonant financial services product can go a long way.

4) What capabilities do enterprises need to build to be successful in AI and how can they accelerate AI adoption?

Karl from Solstice noted that in many cases, technology today is evolving faster than we can adapt to it, both from a consumer use-case perspective and from an implementation perspective.

As enterprises look for business challenges and pain points for which to pilot new technologies, there can be a desire to dive right into the hairiest ones, assuming that the complexity of the challenge is commensurate to the technology being used to solve it. We’ve seen this time and time again with IBM Watson. In many cases, it’s not that the technology failed per se, but that organizations ran out of the time and patience needed to see the project through.

A better approach is to look for highly impactful or painful circumstances that involve fewer moving parts and variables. Highly repetitive yet single-minded tasks; foundational customer experiences where a little bit of additional insight or personalization can go a long way; and highly transactional experiences causing lots of friction today are great examples.

Second, the panel noted that AI technologies and capabilities underpinning CX that move beyond the MVP stage will be far less effective if they area created in silos. Organizations need to continue to find ways to reach across the aisle to other departments and divisions.

The aim should be to create consistent and scalable experiences based on a cohesive approach across the many interactions in the journey. Intelligence created in one area can be leveraged in another, thus gradually synthesizing corporate intelligence into a single, cohesive, intelligent brand experience. This is the exact opposite of the classical “random acts of chatbots” that are happening in many firms at the moment.               

For example, an intelligent search capability can be leveraged by a customer service chatbot. A smart data capture form, which includes in-line processing for credit risk and eligibility, can also be leveraged by an Alexa skill.

Third, advancing data maturity continues to be a massive opportunity and challenge for many firms. Virtually all advanced machine learning that moves beyond the pilot stage needs a viable set of clean, organized, and, in many cases, structured data from which to work. Virtually all hands went up in the air when the following question was asked: “Who currently has a project underway to improve data quality and consistency in your organization”?

Finally, the point was raised that as this technology starts catching on and humans create algorithms that will eventually make decisions on their behalf, the need to start with a better understanding of the customer from the onset will become even more paramount. Enterprises will need to strive to become more emotionally intelligent, placing greater emphasis on activities like user research and human-centered design.

5) What’s the best way to think and talk about the new relationship between humans and machines?

Whereas much of the popular debate is about net gains and job losses when it comes to AI, the panel took a different tact in discussing how we can leverage the technology to enrich the jobs and lives of employees, customers, buyers, and users.

As an example, Mauro from Narrative Science stated that in financial services he is often asked the question of how a bot can act as a financial adviser. He wondered whether that was really the best question to be asking.

The better question, in his opinion is how product teams can eliminate basic and transactional requests, freeing up more of their current financial advisors’ time and enabling them to better service their clients. He feared that the misinformed debate of technology versus human, especially among organizations with large field teams, will keep many organizations on the sidelines for longer than necessary.

The same was said of customer service agents. The enterprise that is focused on empowering its agents with richer customer and incident information, sentiment analysis, and suggested resolutions will be on the fastest path toward happier customers and a lower average cost per interaction.

In the case of predictive maintenance, the panel raised the issue that no matter how advanced the technology gets in reading individual machines or vehicles, it will never be able to eliminate the external circumstances that often cause breakdowns in the first place. The question, then, is how managers can help their employees get even a step or two ahead of where the next trouble spot will be and help them best allocate their time.

6) Where is AI ultimately headed in regard to enterprises?

A key end state many point to for AI is general intelligence, or a machine with cognitive abilities that mirror those of a human. However, the panel pointed out that as of today, there is no linear path for us to achieve this state. The situation today is not comparable to computing power continuing to double year after year, processing speeds becoming increasingly faster, or computing form factors continuing to shrink. We’re going to need several significant, unpredictable breakthroughs (think discovering electricity) to make this happen.

In the near term, AI-related technologies will continue to become more persuasive to the point where virtually every digital experience and interaction will be enhanced or automated to some degree through AI. The speed of adoption will accelerate in the coming years as technology becomes democratized, largely through the likes of Amazon, Apple, Google, Facebook, and Microsoft. The challenge will then become less about technology and more about each organization’s ability, readiness, and willingness to rethink business as usual.

To learn more about our AI practices, go to Solstice.com/AI

Interested in attending our next event. Drop me at an email at dptak@solstice.com We have a full event schedule planned for 2018 across Chicago, New York, and London.