Artificial Intelligence Adoption. A McKinsey study finds that the pandemic has led to companies’ increased embrace of digital technologies, including artificial intelligence, as well as their familiarity with cloud usage. Fifty-six percent of all respondents say they have adopted AI in at least one function, up from 50% in 2020.
A McKinsey & Company study, based on 1,843 surveys of company representatives from all regions, has analyzed the extent to which artificial intelligence has been implemented in the last year.
Fifty-six percent of respondents say they have adopted AI in at least one function, up from 50 percent in 2020. McKinsey notes that companies around the world have continued to incorporate AI into their products and processes in 2021, with substantial benefits. Not only are they adopting more and more complex practices, but they are using the tools more efficiently and increasingly taking advantage of cloud resources. This year, the use of AI increases more in companies based in emerging economies (from 45% in 2020 to 57%) than in developed economies. At the same time, the proportion of respondents attributing at least 5% of their profits (EBIT) to AI increases by five percentage points, from 22% to 27%.
Although the perception of technology’s contribution to revenue improvement remains stable, in this latest analysis it is attributed a much more significant role in cost reduction. For almost all the areas analyzed, the majority of respondents say that its use has enabled them to reduce costs by more than 20%. According to the firm, it is logical that the application of AI tools implies a significant direct cost reduction effect in the first year, which could be mitigated in subsequent years if the pace of implementation does not continue. In this regard, two-thirds of respondents indicated their companies’ intention to maintain investments in AI over the next three years, as already stated in the study published in 2020.
Best practices, a differentiating factor
Companies that see the greatest impact on the bottom line from adopting the technology are more likely to follow both basic and advanced best practices, which include Machine Learning Operations (MLOps); moving their AI work to the cloud, and spending on AI solutions more efficiently and effectively than their peers.
In the case of core tools, the biggest difference between the two groups of companies (those that use AI a lot and those that use it less) is with respect to having a clear framework for AI governance that encompasses the model development process. This is a practice adopted by 38% of respondents in companies with high AI usage, compared to only 20% for all other companies. Bringing rules and procedures to the development phase often requires involving suppliers and customers, which can be a disincentive for those who are less convinced, despite the advantages of having a solid and comprehensive governance model in place from the outset, especially in processes involving different areas.
In terms of advanced data management practices, around half of the companies that make the most intensive use of data identified with the actions proposed in the study, except in the case of “generating synthetic data to train AI models when there is a lack of sufficient real data”. This practice is limited to 27% of companies, whether they are advanced in the use of AI or not. When it comes to data, the biggest difference is seen when it comes to “having scalable internal processes for labeling AI training data,” which is used by just 22% of non-AI advanced companies, compared to 48% of advanced companies.
The use of advanced models, tools, and technologies is minimal in companies that use AI less. The biggest difference with advanced companies is in “adopting a full lifecycle approach to developing and deploying AI models,” which is used by just 26% of the least advanced companies, compared to 57% of the most advanced. It is also striking that only 16% of the least advanced companies consider “Renewing the range of AI/ML technologies every year to take advantage of the latest technological advances.”
The study shows that these practices help the most advanced companies to industrialize and professionalize their AI work, which generates greater efficiencies and better results while increasing the predictability of the associated costs. Part of the efficiency, for the most advanced companies, comes from intensive use of the cloud.
On the other hand, the study also reminds us that the development and use of AI tools are not without risks that companies must work to mitigate, an area where there is always room for improvement. In fact, the greatest perceived risk continues to be cybersecurity, although equality and justice climb positions compared to last year. Also noteworthy is the difference between the relevance that advanced economies attribute to risks associated with regulatory compliance or the explainability of models, compared to emerging economies. Emerging economies, on the other hand, attribute higher importance than advanced economies to the risks associated with the displacement of workers. Respondents point to the need to prioritize among the risks they address, given the lack of capacity to address them all. However, in advanced economies, they consider that they wait until the law begins to require the prevention of such risks before doing so.
In general, the most advanced companies report using more practices related to testing and data checking, identifying biases in models, or keeping evidence and documentation on their performance. The differences over the other companies are significant in almost all cases, although one exception is worth noting: 24% of advanced companies have legal and risk professionals to work with data scientist teams to help them understand the definitions of bias and protected groups. In less advanced companies, however, this caution is used by 26% of respondents.
¡Stay tuned! Go and visit our blog. We post new articles every week.