Skill Diffusion

How Do Skills Change?

The skills needed to do a job have always changed over time. But now, they're changing faster than ever thanks to the internet and technology. The distributed and permissionless nature of the internet and the explosion of content on it have made it possible for anyone to learn almost anything.

This presents an interesting challenge for incumbents who are seeing their jobs morph in near real-time. How do they stay competitive? The question is one of reducing downside: how do you make sure you don't become obsolete? On the other end of the spectrum, are the challengers to the job. They are rarely direct competitors seeking to eat an incumbent's lunch, but they often end up eating it anyway. Where do they come from?

Creating Differentiation

Differentiation in this context is simply the novel use of skills to do something. It can be a new skill, emergent technology, or a unique combination that hasn't been tried before. Regardless of its origin, it is different to current forms of attempting to accomplish a given task.

The challengers who create these differentiations often are not actively seeking to replace a job; instead they usually appear when seeking to fill a gap, trying to solve a specific problem in which they often don't have direct domain expertise.

Let's look at the differentiation of growth hacking skills for example. Most initial practitioners of growth hacking were not marketing people. They had mixed backgrounds that were partially analytical, partially psychological and often had some programming skills. Under the duress of building a startup or product with a limited budget, they tinkered with tricks, hacks, and triggers to try to get growth. Several of these techniques like SEO, referrals, and social media content became incredibly successful. Growth hacking combined skills that were previously never utilized together like data analysis, social psychology, and programming to create a differentiated approach to acquiring customers.

Diffusion to Incumbents

After differentiation is created, it diffuses first to an initial set of incumbents under the pressure of wanting to accomplish similar tasks. Usually early adopters start experimenting and are aided by the creation of new tools.

The combination of better tools and successful outcomes results in a positive feedback loop that creates a formalized skill base. This formulation - not the initial differentiation - subsequently becomes the "skill" for the incumbents to learn.

Let's look at how growth hacking diffused to marketers. Marketers with pressure to acquire customers started adopting growth techniques. Enterprising entrepreneurs, seeing a gap in the market, started creating tools and software to reduce the technical nature of implementation. These tools reduced the cost of learning for more marketers. The tooling affected the marketers by granting new capabilities and the marketers affected the tooling with their feedback and their dollars. Over a few cycles of this, growth marketing starts to grow up with its own canon of knowledge, techniques, and tools. Mainstream companies start taking note and adding growth marketing skills to their job requirements. A few articles get written about how its the new hot skill for marketers. The narrative solidifies and growth marketing gets added to the must have skills list for marketers.


At the final stage of the diffusion, the new skill gains ground over most of the job domain area. The concentration changes over the dispersion time from extremely high and restricted to a small area to relatively low and spread across a wide area. The "settled" dispersion is one that looks quite different from initial source. Throughout the process, new jobs get created, other jobs get destroyed, and new talent cycles through.

People Replace People

If you zoom out and ignore the process of skill diffusion, it will appear as though technology is replacing people. There is some truth to this, because technology is part of the destructive process of incumbent skills. But it's not the full picture. Technology is also part of the constructive process of creating challenger skills, which eventually disperse to incumbents. When the process settles, the composition of who remains in those jobs will be different.

Ultimately, it is people using new skills who replace people using old skills. The cycle will continue for as long as humans want more.

Machine learning is often seen as a future mega-replacer of people. Like with any skill change, machine learning is not itself replacing people. Rather people who use machine learning are replacing people who don't. The uniqueness of machine learning is its ability to change the scale and speed of this replacement. One person building a machine learning algorithm has the potential to replace a very large number of people who don't - and quickly. This kind of scale and speed is quite different from the ones we are used to. Our current labor market system is capable of handling varying levels of displacement over time. It is debatable if it is capable of handling a large displacement of people at once. Very few systems can handle sudden, extreme shocks well.

Building in better ways to handle the skill diffusion process is beneficial to labor market participants. Stopping the process is neither feasible nor desirable because it is a metabolic function core to the general creative destruction process of the economy. Changes during the process may hurt some individuals, but it also provides new opportunities for others like new knowledge, new companies, and new jobs. In order to absorb a change at the scale that machine learning can apply, it is essential to actively try to create a specific form of skill diffusion that utilizes machine learning to enhance a person's job instead of replacing it.

Being a challenger creating differentiation is clearly better than being an incumbent fighting to remain relevant just as it is better to set the proverbial treadmill speed yourself rather than run at a pace dictated by others. But it's not static. One can be a challenger at one time and an incumbent the next. As long as people have some ability to move between these roles, the ability to obtain opportunities can remain largely accessible.