Enterprise AI: Think Big, Start Small
Q&A With Thomas Davenport
Many large firms are aggressively pursuing AI projects today, but those that succeed tend to be less ambitious “low-hanging fruit.” How can firms maintain a focus on large-scale objectives with small-scale projects? And how can they get beyond pilots and proofs-of-concept to production deployments? Thomas Davenport will describe this balancing act, drawing from his new MIT Press book, The AI Advantage, during his keynote address at the MIT IDE Annual Conference May 23.
Tom is the President’s Distinguished Professor of Information Technology and Management at Babson College, co-founder of the International Institute for Analytics, a Fellow at the MIT Initiative on the Digital Economy, and Senior Advisor to Deloitte Analytics. He teaches analytics/big data in executive programs at Babson, Harvard Business School, and MIT Sloan School and is author of more than a dozen books, mainly on analytics and automation. He will also lead a panel discussion on process automation at the May 22 MIT CIO Symposium.
IDE Content Manager Paula Klein caught up with Tom for a preview of his discussions and to talk about some of the most vexing aspects of corporate AI implementation. Highlights of the conversation are featured here.
Q: Businesses tend to think of new technology on a continuum -- ongoing breakthroughs that they adopt incrementally. Is AI different in what it can achieve and in its capabilities? How should it be viewed by top executives today?
TD: The business value of AI is solid rather than sexy or splashy. AI will improve products and processes and help executives make better-informed decisions —important, but largely invisible tasks rather than entire jobs or processes. Short term, AI technologies will augment human capabilities, with smart machines that work alongside smart people. AI should be used to automate structured and repetitive work; provide extensive analysis of data through machine learning, and engage with customers and employees via chatbots and intelligent agents. Companies should experiment with these technologies and develop their own areas of expertise. They need to acknowledge and address the ethical and workplace issues that can arise from AI and automation.
Not every business is an Amazon or Google, which are well on their way to becoming “AI-first” corporations. Many non-tech firms use AI primarily in one business function, such as marketing. Manufacturing and service operations, sales, and even HR are also common functions for using AI—at least the more straightforward versions of it.
Q: Your recommendations that businesses seek out easy-win AI pilots, and create centers of excellence are practical, but are they enough in these highly competitive times? Do companies need to step up their AI game and make bigger investments?
TD: Even very aggressive companies like Amazon say that the bulk of their AI investments are “quietly but meaningfully improving core operations.” If you are a very sophisticated company in terms of technology, however, it might make sense to have one or two “moon shots,” as Amazon does.
A dedicated AI unit will help entrench AI in the organization and make sure applications and use cases move forward. Having a roadmap for setting up an AI Center of Excellence is important because companies are devoting considerable financial resources to AI, and necessary skills and experience are too rare to scatter around the organization with little coordination or collaboration. Just as e-commerce groups supported online presence and commerce, AI will engender new competence centers, or centers of excellence, and new roles within them.
One of the most critical factors in successfully building an AI center is recruiting, attracting, or building talent. That need can’t be overstated, but there are various ways to accomplish the objective, including retraining existing analytics talent. Most firms don’t use all the talent development tools at their disposal—and some survey results from Deloitte suggest that many executives plan to only hire new talent for jobs affected by AI. I’m not sure that’s going to be feasible or fair to existing employees.
Q: Are non-tech and legacy firms at greater risk if they don’t jump in? Farmer’s Insurance seems like a good example of a firm taking advantage of AI incrementally...can you name another legacy company that’s taking a more aggressive stance?
TD: While many companies struggle with pilots and coordination, Farmer’s Insurance is managing an enterprise-wide rollout of AI. AI-based image recognition is used to process car insurance claims while chatbots and other natural language processing-based tools are used in contact centers. The potential of machine learning algorithms is being explored as part of a usage-based insurance program, called Signal. The company is also using automated machine learning tools to develop better predictive models.
Every company approaches AI differently and with more or less urgency. Deutsche Bank, J.P. Morgan Chase, Capital One, Pfizer, Procter & Gamble, and Anthem are among the non-tech firms that have created hundreds of AI projects and are proceeding aggressively to deploy them throughout their businesses. Some even aspire to make AI the core of their strategies. It may be possible to succeed with a more conservative approach, but it will be difficult to be a “fast follower”—there is just too much data, learning, and organizational adaptation involved.
Q: How should businesses choose and fund pilots and experiments? Are there cases where discreet ‘sandboxes’ might produce more innovative approaches? When do you expect that AI-driven businesses will be the norm and how will it impact jobs?
TD: Most pilots and experiments aren’t terribly expensive and are typically funded out of R&D budgets. The challenge isn’t doing them, it’s in “deployment”—integrating them with production systems and processes. Companies increasingly need to create deployment pipelines to ensure that a high percentage of experiments make it into production. On the jobs front, so far there has been little impact from AI. Production deployments will have greater impact on jobs, but I still think we are looking at more augmentation of human labor than widespread automation of it. Of course, despite all the predictions, nobody knows for sure what will happen to jobs.