A General Take on Specialism in the Looming AI Shakeup

A General Take on Specialism in the Looming AI Shakeup

A jack of all trades and a master of none, though oftentimes better than a master of one.

Over many decades, the corporate world has been distilling and refining its organisational conventions towards ideals like "productivity" and "efficiency", and during this process it has apparently become obsessed with "specialists"... But why?

I've observed a significant shift in the duration of my career, where technologies and disciplines mature, and job titles become more and more specific. When I got started in the late-noughties, a "Web Designer" would design things in Photoshop, then code them in Notepad++. Nowadays, a designer is likely to specialise in Visual Design, UI, UX, DesignOps, or Systems, to name a few. These are all valuable roles, and having specialists to fill them is vitally important, but I feel there is a cost to employing this granular approach in a way that is overly strict, which is often missed.

This trend is far from unique to design; it seemingly spans all facets of corporate operations. Specialists can be easily measured in terms of their abilities and their productivity, since there is only one essential dimension by which to measure them. This also makes them easier to incentivise with narrow KPIs and targets. This fits corporate hierarchies neatly - all the people assigned to a specific area of expertise can be grouped into a team and graded on a curve.

Various articles, whitepapers, and studies have attempted to state the importance of "T-shaped" skill sets for decades, representing both breadth and depth. Even back in 2008, David Ing observed:

In most organizations, T-shaped skills are not created as a deliberate policy but emerge because individuals have been willing to risk a somewhat marginal career. Most formal organizational incentives encourage I-shaped skills — the deep functional experience represented by the T’s stem.

And while I agree with this observation, and can anecdotally support it with my own experiences, the "T-shaped" skill paradigm still implies a single area of specialty, with a breadth of superficial skills. These rigid measures of skill depth are self-fulfilling, in that the development of skills is often fundamentally limited by organisational structures that offer little opportunity for broader learning and development. People are actively incentivised to stay put and hone in on just one thing.

There is, however, a demographic who specialise in learning new skills. They can often pick up new skills quickly, and the more skills they've learned, the more transferrable knowledge they can apply to the acquisition of subsequent skills. These people are not antithetical to specialists; in fact many of them are specialists who have a supplementary palette of other skills that afford them the ability to collaborate well with specialists from various different disciplines.

Understanding the Generalist

Generalists, polymaths, or "The Renaissance Man" are often misunderstood. Having a wider view doesn't compromise depth. Through varied, lengthy experience spanning disciplines, generalists can gain a holistic understanding of how lots of moving parts fit together in the big-picture. Combining this view with some deeper expertise in a few areas provides an emergent quality to their contextual awareness. This is arguably the "secret sauce" behind innovative thinkers such as Leonardo Da Vinci and Archimedes, and it's unfortunate that the title of 'polymath' seems nowadays to either be more associated with tech-bro CEOs than genuine innovators, or to be wrapped up in egocentric posturing rather than a humble pursuit of knowledge.

In the professional world, generalists can identify cross-disciplinary issues that specialists may overlook. They can also spot downstream issues from technical and operational patterns that regular specialists using the systems day-to-day would not notice.

So what gives? Why is it that our society has such trouble recognising this obviously apparent and influential demographic, and what can we do about it?

Rethinking Diversity

Diversity quotas are designed to help address systemic inequities in professional representation. They constitute a significant improvement in corporate hiring practices, especially for gender diversity, and for people from minority ethnic groups. However, automatically putting demographically diverse people into specialist roles by default misses the point to some degree.

Neurodiversity is still often heavily stigmatised. People with autism are often implicitly expected to be savants, and people with ADHD are seen as lazy or disengaged. In fact, people with ADHD are especially likely to have more generalist skill sets.

I propose that the problem here is rigid organisational structures that force talented people into specialist roles that do not accurately represent their skills. Not only does this throttle their personal development, enjoyment, and investment in their work, but it may also encourage them to feel the need to "mask", to protect themselves from discrimination.

True diversity emerges when structures adapt around diversity, rather than simply including it to check a box. Operating with transparency and treating people as adults can afford generalists and diverse thinkers the ability to innovate in ways siloed specialists cannot. Diverse, inclusive teams improve decision-making by considering more perspectives, and boost morale by allowing people to be their authentic selves. This is what diversity is supposed to be about.

The Coming Wave

Let's talk about AI. I pinched this heading from the title of a new book by Mustafa Suleyman who is best known for co-founding DeepMind. I recommend the book, though I don't necessarily agree with his views on certain things, such as the concept of "containment".

Even with the exponential advances in AI capabilities we've seen this year alone, I still don't think most people are able to grok the absurd level of disruption our entire civilisation is about to be subjected to. It's really hard to understate how much our lives will be impacted by advances in technology that are compounded by exponentially rising AI capabilities. The specifics of this shake-up are mostly out of scope for this post, but it is safe to assume that specialised knowledge workers are directly in the firing line, since these jobs tend to be the most highly exposed in terms of AI displacement potential.

AI is the ultimate specialist. Giving an AI model some specific parameters, success metrics, and lots of compute seems to be a recipe to surpass human capability in a narrow field. The big target that most AI labs are racing toward currently is the achievement of "AGI" or Artificial General Intelligence. At present, this is a theoretical notion in which exposition of large models to multimodal training data and generation capabilities may allow a greater level of interdisciplinary success, and holistic decision-making potential, in a way that is at least comparable to a regular human. This may happen extremely quickly, or it may be that there are some other unforeseen hurdles in the way. We still don't know if AGI will be able to exercise intuition or conceptual synthesis in the same way that a human can. If it does, then we're all out a job and we have bigger conversations to have than anything I will raise in this article.

For now, the fact remains that those with broad skills will fare better in the wake of job displacement by highly-specialised AI tools. Moreover, generalists will likely be well equipped to make practical use of these narrow AIs - you could command a fleet of them to perform a range of tasks, all converging on a shared goal, without the need for any individual agent to have knowledge of that goal.

How To Do Better

Once, while I was working at a company that had acquired a smaller business, we were trying to figure out how to integrate people from that business into our department. Most people fit neatly into existing teams, but there was one person, who I will refer to as Alex, who nobody could figure out what to do with. By all accounts, Alex was absolutely indispensable - a problem solver who people came to when they just needed something done, even if they weren't sure exactly what that thing was. Alex seemed so highly regarded, and yet nobody could figure out specifically what Alex actually did.

The problem here was that word: "specifically". Alex didn't do anything specific. Alex did lots of different things. We didn't have a department structure that allowed somebody to exist between teams. Everybody had to report directly to somebody, but each somebody represented a singular function.

In the end, Alex was tasked with various odd jobs, and people started to realise those jobs were being done really well. Within a few months, our entire department echoed the sentiments of those we spoke to originally. Alex could do pretty much whatever you needed, even if that meant learning how to do it first, which also never took very long. Alex was a generalist, and all the difficulty of this transition was caused by the rigidity of the surrounding operational structure.

So how can we do better? It's actually not that hard. Sometimes all it takes is an open-minded shakeup (even if the shakeup itself is artificial and involuntary).

Embrace flexibility in your hierarchies, look out for people who have multiple interests and allow room for them to carve out their own role, rather than simply prescribing them a strict function. More often than not, they'll settle in a place where they can still excel at many of the specialised tasks you initially expected, but can also traverse disciplines in a way that yields unique, emergent qualities.

Occasionally, instead of telling people what they need to work on, it's worth asking them what they want to work on.