There is significant demand for AI/ML practitioners today, with the existing talent pool concentrated at a few companies like Google, Facebook, Amazon, and Microsoft. These companies have often paid significant sums of money to attract and retain strong AI and ML talent. For example, Google’s 2014 acquisition of DeepMind Technologies – and its 75 employees – for an estimated $500 million averaged out to more than $6 million per employee.
Often, data scientists at businesses tend to be analysts or “citizen data scientists” who are typically trained in business intelligence and data analysis, but who may have some AI/ML experience. Or they may be statisticians who are trained in sampling from data sets to draw inferences. Most companies are just starting their evolution towards AI and do not have a strong bench of AI specialists.
However, there seems to be emerging recognition that strong, technical AI/ML talent will be important in many industries. On sites like Glassdoor and LinkedIn, machine learning engineers, data scientists, and big data developers are among the most popular jobs. These job postings require candidates with specialized backgrounds that include both advanced math and software expertise.
Given the strategic business value that stands to be captured from AI/ML and the critical importance of these technologies in providing competitive advantage, large enterprises should plan to develop in-house AI/ML expertise.
Many organizations ask us about leveraging citizen data scientists – analysts who have been trained in some AI/ML techniques – to power their AI transformations. But based on our experience and given the significant challenges involved in applying AI/ML algorithms – including challenges with framing the problem, correlation vs. causality, bias in datasets, and information leakage – it is difficult for an enterprise to achieve its AI transformation without a strong, core, technical AI/ML team. This technical team will be central to unlocking disproportionate economic value from the most complex applications. In our opinion, the science of AI is still too early in its development cycle to be entrusted to a non-technical team of citizen scientists and developers.
Nonetheless, we see a strong need for citizen data scientists to support and unlock significant value from a long tail of less complex AI problems. But we recommend that these citizen scientists be complemented and supported by a core, seasoned technical team.
The following figure illustrates the concept:
Figure 41 At most enterprises, there are a few AI applications with significant complexity and value plus a long tail of other use cases with smaller value pools.
The core technical AI/ML team can author and publish advanced algorithms and services that can be leveraged by citizen scientists. This team can also review algorithms published by citizen scientists to ensure they are robust prior to their inclusion in critical business processes and decisions.
Given the significant increase in technical data science and AI/ML academic programs over the past decade, it is feasible for companies to attract and retain this talent. The rise in well-paying data science jobs has caused a surge in enrollment in data science programs: graduates with degrees in data science and analytics grew by 7.5 percent from 2010 to 2015, outpacing other degrees, which grew only 2.4 percent. Today, more than 120 master’s programs and 100 business analytics programs are available in the U.S.
There are also a growing number of data science boot camps and training programs for aspiring data scientists. These programs take in professionals with strong fundamental mathematical backgrounds, including mathematics, physics, or other engineering disciplines, and prepare them for careers in AI. Some of these boot camp courses are available online. Coursera, for example, offers online curricula for both machine learning and for deep learning. Other courses are in-person, such as the Insight Data Science program in the San Francisco Bay Area.
We typically recommend that companies seek to hire and develop their AI/ML talent from academic programs, with potentially a few lateral senior hires to build out the team. Many of our clients already have active university partnerships and programs in place. But these programs are usually not focused on technical AI/ML talent. Small changes in engagement and focus for existing university programs can result in significant improvements to an enterprise’s AI/ML technical team.
Internal recruiting can also play a key role. Many of our clients already have individuals with the right technical profiles, but they often are dispersed across a wide range of internal teams and departments. We often help our clients run internal recruiting campaigns using tools such as LinkedIn to recruit and consolidate their existing talent into AI/ML Centers of Excellence (COEs) that can unlock disproportionate value for the enterprise.
In our own AI/ML recruiting, we have learned that, rather than looking for individuals with skills in specific techniques, we should select candidates with strong technical skills who also have strong mathematical foundations and intrinsic problem-solving skills that show their potential for learning a wide range of algorithmic techniques. In general, algorithmic techniques can be coached and learned over time – and are constantly evolving anyway – but mathematical fundamentals are much harder to learn.
The following figure summarizes the background and experience we recommend as part of building a core AI/ML team.
Figure 42 Typical required skillset for AI/ML practitioners