About seven or eight years ago I realized that the analytics and big data research, writing, and teaching I was doing was going to turn into AI research, writing, and teaching. I could have been earlier to come to this awareness, but at least I wasn’t totally clueless about it. But AI itself keeps changing. Over that time we’ve seen a lot of evolution in how enterprises think about AI as a business resource. I was motivated to reflect on this topic by the release of the 2020 Deloitte State of Enterprise AI survey, which is available here. I didn’t work on the survey this year but I am a Senior Advisor to Deloitte’s Analytics and AI practice.
There are many different aspects of the survey, but the one that stood out to me was that the respondents feel that AI is getting easier, and will continue to do so. The respondents are 2727 global executives from nine countries, and their organizations have all adopted AI. That in itself suggests that AI is becoming more pervasive, because in earlier years of the survey we had to interview lots of executives in order to find adopters. And the average respondent is becoming more sophisticated and experienced with AI. This year only 27% were “starters”—companies with low expertise in AI and fewer than five production implementations of AI—which is also a lower percentage than in the past (36% in 2018).
Buy, Don’t Build
The single most interesting result of the survey to me is the preference for buying ready-made AI technology over building it. Indeed, at some point it will be difficult not to buy. 74% of these executives agreed that “AI will be integrated into all enterprise applications within three years.” In terms of today’s practice, 50% say they will either “buy all” of their AI capabilities or “buy more than build;” only 8% will “build all” and 13% will “build more than buy.” This is not just an approach particularly favored by novices at AI; the more experienced a company was in the survey, the more likely it is to prefer buying rather than building. Perhaps not surprisingly given the buy-over-build preference, the respondents expect basically universal adoption of key AI technologies—machine learning (97% either using it now or within a year), deep learning (95%), natural language processing (94%) and computer vision (94%)—really soon. I hope that the fact that the respondents were surveyed in December of 2019—pre-COVID, in other words—will not deter their optimism.
There is lots of positive news here if the respondents are correct. They’ll be able to source sophisticated Ai technologies without the hassles of hiring hard-to-find data scientists who can program in PyTorch or TensorFlow. The data they need for AI models will presumably be present within the transactional applications that incorporate AI capabilities. Users of these systems may even find it relatively easy to adapt to these AI applications, since they are already using the systems in which they reside.
Being a Good Buyer and Risk Manager
Of course, there is a less optimistic side of easier-to-use AI. If firms are going to buy their AI applications rather than build them, that means they will need to understand the vendor marketplace for AI offerings and be able to choose strategic partners. But less than half of the AI adopters (47%) say that they have a high level of skill around selecting AI technologies and technology suppliers. They’ve clearly got to bulk up on this particular skillset.
With development of AI applications being less of an issue, how are these companies planning to establish a competitive advantage with AI? No one approach appears to dominate, but “modernizing our data infrastructure for AI” is the most popular response, with 20%. Gaining access to the newest and best AI technologies, moving to the cloud, deploying data science platforms, and developing partnerships for fast execution were also relatively popular tactics.
There are also a number of risks that these AI adopters appear to be concerned about. We first began to ask respondents about these in the 2018 survey, and they have appeared consistently since then. This year, 56% agree that their organization is slowing its adoption of AI technologies because of the emerging risks; the same percentage is worried about a backlash by the public about AI. Some of the specific risks that most concerned respondents were cybersecurity issues, AI failures that might affect business operations, misuse of personal data, and regulatory changes involving AI.
Regardless of these risk concerns, those who have adopted AI are quite bullish about its value to their businesses. This finding has appeared in the survey data every year it’s been conducted. In this year’s survey, 73% of adopters say AI is very or critically important to their business success today, and 71% expect to increase their investment next fiscal year.
This latest survey certainly indicates that AI is becoming more pervasive and easier to adopt, and that companies are benefitting from it. A few risks from AI need to be managed, but they shouldn’t hold any company back from applying AI to improve their business. The survey results suggest that companies are surging forward with this technology.Share This!