The search and competition for talent between organizations is typically dictated by the type of problems they face. If those problems require more vertical knowledge and creativity, they will look for specialists. If they require more horizontal knowledge and creativity, they will look for generalists.
The metaphor of knowledge glasses
I like to think about knowledge as multiple glasses that can be filled. The total number of glasses represents knowledge areas horizontally. The filling level of each glass represents the depth of knowledge in each area.
Following this logic, a specialist tends to have two or three very full glasses, while a generalist tends to have more partially filled glasses.
Thinking now about the implications this has on the mode of operation
If we consider creativity as the ability to recognize patterns and connect different points of knowledge, then the specialist will tend to be more creative inside their verticals, while the generalist will tend to have a more horizontal creativity.
Combining this creativity with taste (1), meaning judgment ability, we end up with very different operating modes between both profiles, since both tend to create and select very different paths to solve the same problem.
Let us look at a small mathematical example regarding both profiles.
The Specialist has 2 glasses with 10 levels of knowledge each.
The Generalist has 5 glasses with 4 levels of knowledge each.
a: number of glasses.
p: filling level of each glass.
| Metric | General formula | Specialist | Generalist |
|---|---|---|---|
| Total knowledge | a × p | 20 | 20 |
| Combinations inside similar areas | a × C(p, 2) | 90 | 30 |
| Combinations between different areas | C(a, 2) × p2 | 100 | 160 |
| Depth / horizontality ratio | p / a | 5 | 0.8 |
| % interdisciplinary | C(a, 2) × p2 C(a × p, 2) | 52.63% | 84.21% |
| % inside the same area | a × C(p, 2) C(a × p, 2) | 47.37% | 15.79% |
Naturally, the boundary separating a specialist from a generalist is difficult to define objectively. In practice, humans tend to identify the specialist profile more easily.
As expected, even with the same total amount of knowledge, the generalist profile creates more possibilities of combination between different areas, while the specialist profile concentrates more combinations inside the same area.
Ideally there would be a profile with all glasses completely full, but humanity is not built for that. Our best chance of approaching that scenario comes from the combination between different profiles.
Imagine a tribe with 150 people. Each person knows how to solve certain problems. Some will have more vertical creativity, others more horizontal creativity. Alone, one person might survive, but together, the tribe can solve far more problems because knowledge starts combining.
"If I know 10 ways to build traps and another person knows 10 good places to put them, together we have 100 possible combinations. If another person knows how to preserve meat, the number of possibilities continues to grow so we can survive the winter (...)" (2).
This creates a problem, because these combinations only happen if there is connection between people. If I do not know that someone possesses a certain type of knowledge, I will never be able to combine it with mine (3). It is then up to organizations to understand, connect and create environments that maximize these combinations.
What impact will AI have on these profiles?
In my eyes AI is clearly creating a convergence between specialists and generalists. A specialist can now cover adjacent domains without needing external help. And a generalist with strong prompting ability can now produce work that previously required a specialist (4).
At the end of the day, we are reducing the cost of competence and that makes execution work, which previously required deep collaboration between multiple functions, increasingly faster and cheaper.
Following this logic, the specialist becomes fundamental because of deep judgment ability. Through experience and, mainly, through the ability to understand when AI is wrong inside their domain.
The generalist, on the other hand, will create value through system orchestration. To do that, they will need to properly deconstruct problems across different domains so systems can solve them.
The truth is that, if we think carefully, AI will not fill anyone's glass, it only gives us access to the remaining glasses. Until now I believed that in the future there might exist a profile resulting from the combination between specialist and generalist, but the truth is that this is simply the result of AI access tending toward infinity. For organizations, this means that whoever does not build systems capable of combining knowledge through AI leverage will fall behind in the race.
Feedback is welcome as usual :)
(1) Virgílio Bento, Some Thoughts on Taste
(2) Pedro Domingos, Can AI Help Build a Collective Human Intelligence?
(3) During one of my internships, they searched for external help to solve a problem simply because they did not know I already had the required knowledge.
(4) Lenny's Podcast, How we restructured Airtable's entire org for AI