C-Suite and Board-level Thinking on AI Needs a Reset

It has been well-documented in numerous recent surveys that companies are finding it hard to adopt AI because they lack employees with the diverse set of required skills and don’t really know how or where to start. Of the companies who somehow manage to overcome these initial barriers, a majority have struggled to move their pilots to at-scale deployments. As a result, their AI-driven transformation takes longer, costs more, and delivers ROI slower than their expectations. From my own experience helping companies adopt AI over the last couple of years, I think that many of these problems arise because, broadly speaking, executives in the C-Suite and on Boards don't yet have a good general (high-level) understanding of AI – and what it can and cannot do today.

AI is a general purpose technology (GPT) just like the steam engine and electricity. Economists define a GPT as a technology with many potential applications, which has quite a bit of scope for improvement even after it gets deployed in its initial applications, and which eventually ends up being widely used across a variety of sectors. Research on adoption curves of the steam engine and electricity across the economy has demonstrated that it takes time and work for human beings to figure out how to use a general purpose technology effectively. When electricity was first introduced in American manufacturing over a hundred years ago, factories used to be powered by steam engines. Despite electricity clearly being a cheaper source of energy than the steam engine, it took several decades for manufacturing companies to collectively understand that electricity enabled factories to be designed in a different manner, which made them far more efficient, safer, cleaner and hence much more competitive. 

In contrast to the several decades it took for the economy to transition to electricity, the adoption of AI across a variety of industries may happen faster because of several reasons, for e.g., the amount of money that has been invested in AI startups (~$10B in 2018 alone) enables rapid experimentation with use cases and business models, massive investments by leading AI-first companies like Amazon, Google and others, wide availability of open source tools and frameworks, and best practices spreading much faster now than they did a century ago - thanks to the Internet, which is another general purpose technology. It is thus crucial for executives in the C-Suite and on Boards to start approaching AI with the right mindset and expectations. Here are three practical suggestions:

1. Get Educated: C-Suite & Boards have a responsibility to get generally educated (even if at a high level) on what AI is, and what it can and cannot do today (an executive reading list is provided below.) Increase your understanding of the use cases related to AI adoption in your industry and in allied industries. Without developing this basic understanding, it is practically impossible to craft any semblance of an AI strategy, and the fledgling AI initiatives at your company are likely to run into the same roadblocks described above.
2. Think Strategically: A general purpose technology reduces the cost of some key input, for e.g., electricity reduced the cost of energy and provided much more flexibility in designing factories compared to the steam engine, computers reduced the cost of arithmetic, and AI has reduced the cost of prediction. Think about the different aspects of your company's operations, products and services where you can take advantage of the reduced cost of prediction to strengthen your current sources of competitive advantage or create new ones. And what might happen if your competitors figure these out before you do?
3. Expect R&D-like time horizons: Given the dynamics of a general purpose technology in its early-stages of market adoption, it might be helpful for the C-Suite and Boards to view the building of AI capabilities inside a company (which includes carefully investigating solutions that are available on the market, if that makes more sense to solve a specific problem) as similar to establishing an R&D program - not one for conducting “blue sky research", but one that is closely tied to achieving medium-term (1-3 years) business outcomes. Using AI in the context of solving a specific business problem involves several elements of discovery and iteration, so the experience feels closer to embarking on a short to medium-term R&D project than to, say, a well-planned technology roll out.

Building internal AI capabilities is not a quick-fix to help you deliver the numbers you’ve promised Wall Street this quarter or the next, but an important investment in your mid- to long-term viability as a company. If thought through carefully and executed well, you can generally expect to see ROI from your first AI deployment within 6-12 months. You can then build on that initial success to make good progress over 2-3 years (and beyond) as you continue integrating AI across your operations. And for those of you wondering what happened to the manufacturing companies trying to adapt to the introduction of electricity over a century ago, almost 40% of the incumbents didn’t survive that transition from steam engine to electricity – they either went out of business or were acquired by nimbler competitors.

***

Executive Reading List

1. Prediction Machines: The Simple Economics of Artificial Intelligence by Ajay Agrawal, Joshua Gans and Avi Goldfarb
2. Artificial Intelligence and Machine Learning for Business: A No-Nonsense Guide to Data Driven Technologies (3rd Edition) by Steven Finlay 
3. Applied Artificial Intelligence: A Handbook For Business Leaders by Mariya Yao, Adelyn Zhou and Marlene Jia