Most CEOs recognise that artificial intelligence (AI) will be critical to their business, but few enterprises have actually undertaken AI initiatives so far. A recent Gartner survey of CIOs found that only four percent of respondents had deployed AI in their organisations.
Nevertheless, growth could be dramatic. The survey also found that one-fifth of CIOs are piloting AI or planning to in the short-term. While AI faces the usual obstacles to the progress of unproven and unfamiliar technology, prepare for its inevitable breakthrough and plan strategically.
The technology has tremendous potential, but such an immature technology also poses severe risks for those investing in it. Reduce the risks and increase the value of investments by getting corporate leaders to think about AI in a longer time frame.
If you think about the automobile, for example, it’s a simple idea, but the world had to be moulded around it before we got the full benefits. It took 250 years for the automobile to develop from idea to the car-oriented life we lead today.
AI is another idea that will eventually develop into a general purpose technology (GPT). To reap the full benefits, think in a time frame similar to the automobile. The idea of AI arose about 60 to 70 years ago, and will probably take another 60 to 70 years to be woven into the fabric of modern life.
There are already early successes, such as UK online grocer Ocado, which envisions a delivery chain that’s fully automated by AI. Swarms of warehouse robots bring items to human pickers to place in baskets. Then, another AI-based system of conveyors directs the baskets to trucks for delivery. It takes 10 minutes for items to go from where they’re stored to the truck. Ocado’s Birmingham warehouse now moves 1.3 million items per day.
There are, however, a complex set of processes involved with AI that will surely lead to a few dead ends and false starts before it’s widely adopted by society and becomes a mature GPT.
Reality check no. 1 – immaturity
Given the GPT adoption time scale, AI's maturity needs to be dated at its proper age relative to other information and communication technologies – which is 1958 from a GPT perspective.
AI is still immature across key dimensions of its future function and capabilities. It’s heavily focused on core generic technologies like graphical processing units, deep neural networks and natural-language processing. As it develops, a rapid set of technical improvement cycles will occur along with a pervasive diffusion of the core technology, producing a vast array of everyday products and services.
Many AI systems today require custom, hand crafted development. Few, if any, reliable standards have been established and there are limited engineering cookbooks. We’re in the pioneering era. Development and testing cycles are lengthy. Dead ends and false starts slow the march toward production systems.
Over the next 75 years, there will be multiple generations of the AI industry. A wide variety of spillover effects will spawn new uses, markets and possibly industries. We’re already seeing signs of complementary innovation being triggered. R&D is driving the exploration of new value propositions at accelerating rates in downstream sectors, including financial services, healthcare and mobile commerce.
Reality check No. 2 – plan for the long-term
Be careful of hyperlocal planning time horizons. Your strategic planning process today needs to have some forward-looking perspective for the long journey ahead.
It sounds like a frivolous question, but you’ll waste more money and miss more opportunities today unless you ask, "What will our enterprise's use of AI look like in another 60 years?"
Why is it hard to plan on long time scales? Because most enterprise strategic planning processes occur on time scales that are hyperlocal relative to a 75-year cycle. This view is often driven by three to five year strategic business plans, two year budgeting cycles and quarterly financial results.
Reality check no. 3 – adjust your processes
Strategic planning processes need to be deliberately stretched to avoid hyperlocal planning horizons that may negatively affect R&D, business planning and investment decisions.
Zoom out your strategic planning lens to 7 to 10 years. This extended planning horizon may provide enough forward looking knowledge to help avoid common pitfalls. These include failure to build required competencies; overinvestment in short-lived products and architectures; opportunity costs of flawed decisions or investments; and inability to detect and interpret important technical, social, economic and political events that will determine the course of AI emergence as a GPT over time.
Plan for long-range potential
AI can solve business problems today if you develop applications carefully. But don't let the success you have with them deceive you. In the long run, your enterprise will get more value out of what you learn about AI than from the uses it addresses today.
That’s because the more of a GPT that AI becomes, the more obsolete today's applications, uses and AI-driven business models will become.
To keep your enterprise from becoming trapped in an AI deadend, experiment and plan for the long-range potential of AI applications. Have some vision for where AI could go to make good bets about which investments today will build toward the future.
Leigh McMullen is a research vice president at Gartner. He provides CIOs with insight on navigating and making a difference within the C-suite, particularly around IT leadership, politics, influence, culture, business engagement, internal selling and IT marketing. Leigh will be presenting at Gartner Symposium/ITxpo on the Gold Coast, from 29 October-1 November 2018.