8 ways to effectively advocate for AI

Digital business transformation trailblazers who are actively promoting the adoption and use of artificial intelligence (AI) clearly need to be better at communicating, separating fact from science fiction. AI business champions must keep it real and stay focused on tangible business value. A data analytics, big data and AI pioneer shares how business leaders can effectively advocate for AI.
Written by Vala Afshar, Contributing Writer

"The most powerful person in the world is the story teller. The storyteller sets the vision, values and agenda of an entire generation that is to come." -- Steve Jobs

There are not many technologies that elicit responses as strong and emotional as AI. It is presented less as technology and more as an unknowable, mysterious, disruptive force. Headlines in the media stoke the fire by quoting the biggest names in tech referring to AI as something to fear. A recent debate between Alibaba Group chairman and founder Jack Ma and Telsa CEO Elon Musk best exemplifies the varying points of view on AI. So it is entirely understandable as to why people are afraid when AI comes to their organizations. 

Research shows that all lines-of-business are poised for an AI revolution, including customer service, digital commerce, product management, and marketing. To better understand how business trailblazers can effectively position and champion adoption of artificial intelligence (AI) throughout their organization, I connected with an analytics, big data, and AI pioneer who has advocated and implemented AI technologies throughout his illustrious career. 


Ketan Karkhanis, SVP & GM of Analytics Cloud at Salesforce

Ketan Karkhanis serves as senior vice president and general manager for Salesforce Analytics, where he is responsible for driving all aspects of the analytics business, including product strategy, management, marketing, engineering, and distribution. Under his leadership, Salesforce Analytics has become a complete solution to make analytics more accessible to every business user, offering the full spectrum of basic, advanced, and AI-powered analytics. 

Karkhanis understands and is also motivated to ensure that businesses fully understand the complexity, benefits, and best practices associated with the adoption of AI and digital business transformation toward cultivating a data-driven culture aimed at improving the stakeholder (employee, customer, business partner, and communities) experience. 

"Those of us who are promoters of AI clearly need to be better at communicating, separating fact from science fiction. As someone who is out there every day talking about AI, I've developed a personal checklist that I use to keep it real and stay focused on tangible business value. I hope it empowers you to become an advocate for AI in your organization," said Karkhanis. 
Here are eight effective ways to advocate for AI, according to Karkhanis:  

  1. Don't adopt a systems view of the world: The way you bring about change is not by going on and on about systems, but by focusing on the customer's point of view. Instead of technical jargon, speak the language of those who work every day with the business processes changing. In the end, the business value is the metric that matters.
  2. Deliver quick victories: To compose a change management symphony, leaders need to think about people first. Change management is hard not because change is hard, but because it takes so long. Inspire your organization, and most importantly, its people, by delivering rapid-fire micro changes. Small shifts that impact and improve outcomes for a small group of people get talked about and serve as powerful catalysts for future acceptance across the organization.
  3. edecz1wwsaaw2ec.jpg

    What is Artificial Intelligence? 

  4. Be humble: Hype may be the biggest PR problem faced by AI. AI is not a magic wand that, abracadabra, will solve world hunger, bring universal peace, and guarantee a Super Bowl victory. It is advanced mathematics. Period. AI is a tool -- a contributor to finding solutions. People make use of AI over a period of time to dig for valuable gold nuggets and make incremental enhancements to processes and outcomes. It is also important to broadly educate all stakeholders with foundational knowledge regarding various implementations of AI, starting with answering the question: 'What is AI?
  5. Be emphatic: "AI is going to take my job" is a real concern for employees. It may be unfounded, but that is the fault of widespread misinformation. The effective communicator will work hard to reduce those fears by helping people to understand that AI is there to help them to be more effective, not to supplant them. AI is just a math tool, and human beings are very much still its boss. As a voice for transformation within your organization, it is important to empathize with the team and individuals who are going to consume and implement the output of your work.
    Similarly, you don't approach a VP of customer service with 30 years' experience declaring that you "can fix their flagging CSAT score." Instead of leading with your conclusion, engage in dialogue that respects their role. It can be as simple as asking, "Is CSAT something you want to focus on?" As soon as one presents AI-derived numbers as better than someone's vast experience, the technology transformation is heading down the road to failure. Instead, present your insights as an opportunity for them to consider and discuss.
  6. Explainable AI Is equally, if not more important, than accuracy: Part of the fear of AI stems from its perceived "black box" inscrutability. That is why it is so important to employ technologies that present automated discovery as natural language explanations of why something is happening and why a given prediction is important. AI may predict a number, but that number on its own is useless. Lasting value is derived when that number is accompanied by a natural language explanation that allows the business user to understand why the model arrived at that number.
    Business people are not alone in this. Data scientists, compliance experts, and other technologists need their own explanations in the form of the code for the model that delivers the prediction. Such transparency enables them to validate the model and is the best way to engender trust.
  7. Don't boil the data ocean: Among the most common objections to AI are misconceptions about data that start with "We don't have the data for that" and end at "Our data is a mess." This is the legacy of big data hype. And it's wrong. Deep data, the progressive notion that one simply needs the right data for the given use case, is liberating. It suggests that your organization can do this right now. Think about it. Why would one need a data lake with HR data to deal with customer attrition? Service information, customer history information, and order history may be all that's needed to develop valuable insights. This is how to end the wait for perfect data and make projects manageable and doable.
  8. Solve the "it's too…" problem: Too hard. Too complicated. Too expensive. AI has moved forward very quickly. It's no longer the sole domain of four rare geniuses. AI-augmented analytics mean that existing business and data analysts can up-skill to become citizen data scientists. For example, Salesforce Trailhead learning modules provide an easy learning approach that enables employees to take advantage of today's self-service, point-and-click style discovery tools.
  9. Guard against bias: Bias is a very real issue against which organizations must be vigilant. Education is, of course, the first and most important action. Beyond that, though, a combination of governance and technology offers a way forward. Governance means having a clear policy about what data elements will be used in AI models. Technology can then help guard against subconscious bias, the unintentional small ways that human beings stray from the correct lane. Bill Gates was on the right path when he said, "Automation applied to an inefficient operation will magnify the inefficiency." In other words, automated discovery built on biased models will increase bias. Someone using Einstein Analytics to build a model for discounted loan applications may do the right thing in terms of governance and mark "race" as a protected field to avoid its use in decision-making. That's a good start, but not enough. So, Einstein Discovery will take the added step of looking through data, flagging ZIP code as highly correlated to race, and suggest removing it from the model to reduce subconscious bias. 

The biggest hurdle for communicators to overcome is the sensational hype around AI. It is most effective when used for comprehensible, realistic projects. Don't think about AI as one massive, disruptive change in the organization. Its brilliance lies in small moments, augmented by human intelligence, that drive better decision-making. It is not the pot of gold at the end of the rainbow, its the nuggets of gold you pick up along the way.

This article was co-authored by Ketan Karkhanis, SVP and GM of Analytics Cloud at Salesforce.

Editorial standards