500 million members. 20 million companies. 15 million jobs. 50 thousand skills. 60 thousand schools. 24 languages. At LinkedIn, they like to say that Artificial Intelligence is like oxygen — it powers everything they do. And when one looks at the numbers that represent the massive scale of what they are trying to achieve, it becomes evident that LinkedIn would suffocate without it.
Recently at a conference hosted at MIT on AI & the Future of Work, former CEO of Google Eric Schmidt predicted the next wave of successful companies in the global economy will be those that are “AI-first” — rather than attempting to inject AI narrowly into a small set of products, these companies will create an AI ecosystem that impacts every product and service that they provide. LinkedIn is a perfect example of a company that has created such an ecosystem, with machine learning and other AI components influencing every aspect of their operations, both internally and externally.
At the core of LinkedIn’s business is the desire to serve their customer the best possible experience, regardless of the outcome they are seeking — finding a job, building a network, crafting a recruiting list, or sourcing clients. The vast majority of these outcomes rely on LinkedIn’s ability to sift through millions of data points and generate high-quality recommendations. As such, all of LinkedIn’s products and features are powered by machine learning because its algorithms are able to not just follow pre-programmed instructions about how to process such large data sets, but also to learn over time how to increase the quality of its recommendations.
LinkedIn’s initial quest as it relates to machine learning and other AI components has largely been to develop algorithms that master the nuances within each data “dimension” (members, companies, jobs, skills, etc). Within each dimension, LinkedIn’s algorithms try to develop some sense of a “hierarchy” — that is, how data points within that dimension are related to each other (for example, which job seekers are similar to each other). Once hierarchies are established, algorithms can then be developed around how data points within one dimension relate to data points within another dimension (for example, how does a certain set of skills on someone’s profile set them up for success in a certain set of jobs).
While these foundational algorithms are quite powerful on their own, the tremendous capabilities of LinkedIn’s machine learning become even more clear when considering their longer term goal — developing and accelerating insights from their “Economic Graph”, a digital representation of the entire global economy. LinkedIn’s CEO Jeff Weiner has articulated his view that this product will develop significantly over the next decade in a way that will create opportunities to identify more comprehensive trends and insights related to the economy, from upward mobility pathways to talent migration dynamics by region.
If LinkedIn is able to successfully scale the depth and breadth of its machine learning algorithms, there are a massive amount of additional market opportunities that the company will be in a position to tackle — particularly as it relates to diagnosing and prescribing solutions for skills gaps in the global economy. Coincidentally, the AI components that are so essential to LinkedIn’s product offerings are the same technologies that threaten to disrupt the current landscape of the workforce: a study by the McKinsey Global Institute in 2017 found that 50% of all activities in the global economy could be automated using technologies that are already in existence today. Such disruption to the workforce suggests a considerable likelihood that millions of people will need to learn new skills — and until they do, there will be large discrepancies in the supply and demand of skills within individual regions. The real opportunity for LinkedIn is not only identifying the skills gaps that exist as the workforce is disrupted, but to become a central facilitator of alignment between employers, job seekers, and learning programs and to develop agility in this alignment as the pace of change in the workforce becomes even more rapid.
Despite what seems to be a clear alignment between the emerging challenges in the economy and LinkedIn’s growing machine learning capabilities to address those challenges, there are numerous open questions on how to effectively implement such a solution. The most crucial question is around financing: if the most important activity is the upskilling of an individual with a specific set of skills that will prepare them for a specific job — who is paying for that upskilling? The job seeker? The employer? The taxpayer? An investor who shares future income? There are plenty of innovative possibilities that can be considered, but it’s imperative that we begin to lay the foundations for those possibilities now, as they will be essential to quickly deploy as the pace of change in the workforce continues to accelerate.
 McKinsey Global Institute. 2018. “How Can Business Leaders Make The New World Of Work Better For People?”. Podcast. The New World Of Work.
 Schmidt, Eric. 2018. “AI & The Future Of Work”. Conference, MIT.
 Agarwal, Deepak. 2018. “An Introduction To AI At Linkedin”. Linkedin Engineering Blog. https://engineering.linkedin.com/blog/2018/10/an-introduction-to-ai-at-linkedin.
 McKinsey Global Institute. 2017. “Jobs Lost, Jobs Gained: Workforce Transitions In A Time Of Automation.”