Can AI Make Us Great Beginners at Everything?

Most people, intuitively, think that for AI to be helpful, it should help us improve how we do things we are good at. I challenged that assumption in two previous posts (here and here), and it is time to expand on it with additional data. 

First, we do not always perform at our best. AI could support us for tasks where we have the requisite skills, but for which our attention and effort dwindle because they are repetitive or simply because we may not be willing or able to expend much effort on them (think about reading a 30-page briefing document before a meeting).

Second, we are not that good at many things we do. AI can augment us when our skill levels are lower. The importance of this point is often overlooked: all of us, in our daily jobs, from frontline managers to CEOs, spend much time on things for which we are mediocre. For many of us, being a better beginner to perform better in areas we haven't mastered is a foundational part of our professional value. There is ground to believe that generative AI can be helpful there. If this is interesting, read on.

A recent study from the Boston Consulting Group empirically showed that professionals benefit from generative AI when confronted with tasks requiring skills they don’t master. In this specific example, consultants who didn't have data science capabilities could do some data science work with the help of generative AI. 

Those who use Gen AI daily probably already know that from direct experience. Some personal, anecdotal examples:

1. Interacting with doctors for benign but long-standing issues that no individual specialist had bothered (or could) look at end-to-end and for which the family doctor needed more data or the time (or the skills) to connect the dots. Think of the interplay between body parts where doctors would deliver a fully effective diagnosis if they were operating as a team, but given their time constraints, they don’t. Working with tools like ChatGPT, Claude, Perplexity, Consensus, etc, allows to form additional hypotheses and helps doctors to generate more holistic treatments
2. Engaging with lawyers and public officials in a new country in their language where you are less fluent. The result is sounding confident and precise, which matters to be taken seriously, understanding the nuances more thoroughly, and generally being able to cover more regulations faster. (Tools used: ChatGPT, Perplexity)
3. Taking cross-disciplinary approaches to research and analysis by connecting dots between themes for which one's knowledge was limited. This applies to both academic research and enterprise innovation, among others. (Tools used: ChatGPT, Claude, Scholarcy, Perplexity, Consensus, Elicit)

Some of these could have been done in the past with “conventional means”: search, online translation, etc. However, the speed and breadth offered by the new tools allow us to do much more in the same amount of time, radically changing what one can tackle confidently - and changing how we work.

You are your S-curves skills portfolio

Let’s now generalize what this all means. 

Our skills and capabilities are crucial to our performance in specific tasks. With effort, strategy, and luck, we reach higher levels in a few chosen skills over time. Typically, only the initial part of skill acquisition is easy (hence the "S-curve" shape); and some skills decay, some are not our forte, and we continue to struggle.

Much of our activities require us to tap into our less-developed skills, like a football player with a less-favored foot (or hand) and field position who needs to operate within an extensive range. In other words, the skills we bring to bear look like a portfolio of S-shaped curves, and our effectiveness often depends on their average - and sometimes on the least-developed ones (see chart below).

Depending on who you are or the role you have, you need a specific portfolio:

• Generalists, like general managers or disruptive-innovation teams, often operate in skill quadrants where they struggle. That’s where they delegate or bring teams of experts together, incurring additional “transaction costs” intuiting that investment pays off.
• Specialists, like university professors and academic researchers or R&D experts, typically try to spend much of their time in areas where they are world-class. However, they do have to do things that they do less well (say, navigating new types of grant application procedures). Also, in jobs where strong expertise drives most of the success, there might be a need to absorb new knowledge, develop situational awareness, and constantly scan the environment for changes. 

Of course, the above is an imprecise categorization, but it gives us an idea of the possible extreme archetypes. Each of us likely falls in a continuum between those at any time.

If used well, generative AI may make us better beginners and help us travel faster and more effectively rightwards from the left to an "advanced beginner" level.

This can have profound implications. Today, it doesn't take us much time to get the basics of new spaces as long as their knowledge exists publicly or within your firm (among others, management consultants have noticed, and many are building out the next generation of knowledge management tools). These new tools also allow for interplay between us and them, not just to get answers.

If that happens, the portfolio could be significantly improved. Our overall performance is often disproportionally impacted by our weakest spots, and so is our ability to intervene in more areas that require our attention. Learning how to leverage Generative AI to make us dramatically more rounded (in learning lingo, T-shaped, or any other shape than "I") can be a game changer.

The implications of being superhuman beginners

There's a clear "future of work" intuition: traditionally, we were told to stay in our lane, as the transaction costs for acquiring a modicum of new knowledge were high. Experts were tapped into for low-level questions, choking their capacity, and combinatorial innovation (combining ideas from different fields) was very hard. Search engines applied to the internet's fact base changed things. Now, things can change further.

In the longer run, assuming that the tools and the skills to use them improve, this could impact, among others:

• How we manage: We can try to understand things outside of our comfort zone without needing to look for experts’ help all the time or form better hypotheses to test when experts’ limited time becomes available in meetings.
• Workforce planning and talent management: We can use more inexperienced or less qualified workforces as long as they learn how to be effectively augmented.
• How we organize: making smaller organizations, or ones that are less embedded into traditional knowledge ecosystems, competitive.

More generally, does that mean a possible reduction of the size of effective enterprises (a new twist in Coase's law)? More entrepreneurship? Can developing-countries professionals flatten the first part of the curve more easily? Can young people do more than today? Can combinatorial research and innovation get a boost?

We don't know for sure. Much can go wrong. When using Generative AI poorly, dependency and quality control are risks. However, new "augmented thinking" literacy can help, as can a more innovative approach to organizational, process, and work design. Time will tell, but the opportunity seems large.

This article is part of a series on AI-augmented Collective Intelligence and the organizational, process, and skill infrastructure design that delivers the best performance for today's organizations. More here. Get in touch if you want these capabilities to augment your organization's collective intelligence.