AI and the future of knowledge work

Most people have not had a meaningful discussion of knowledge work, we are surrounded by it and it has become the norm and easily ignored (the last thing a fish notices is the water it is surrounded by).

The phrase “knowledge worker” was coined by Peter Drucker in his 1966 book “The Effective Executive”. To make clear the definition imagine a worker whose job it is to bend pipes. This is manual work, you can see the work in front of you, it is “obvious” what the task at hand is and obvious when it is completed. You can easily measure performance (how many pipes were bent) and improve performance by bringing in more people to bend pipes.

Now contrast this with knowledge work. Imagine a strategist tasked with creating a company strategy. You cannot see this work, it is not “obvious” what must be done, measuring the quality of the work is complex, and you cannot bring in more people to increase throughput linearly. Even worse is Brooks’ Law - adding more people to a knowledge work project makes it finish later.

The management of one type of work cannot be used to manage the other. One of the key differences between knowledge work is that it has:

no right answer; there are choices instead - Peter Drucker

Choices though are exhausting - they require rigorous analysis to try to determine a “best choice”. No such rigour is needed when bending a pipe.

The question when managing labour is “How can we do this faster?”, the question for knowledge work is “What is the right thing to do?”. Doing the right thing is what Peter Drucker defined as being effective.

If a pipe bender is not effective this is obvious i.e. poor throughput of completed pipes. For the knowledge worker being effective is much more difficult to measure and understand.

AI

With AI being used to perform the knowledge work, individual contributors are now acting more like managers (of the AI) and are rediscovering knowledge work management techniques.

the task is not given; it has to be determined - Peter Drucker

The specification is key to knowledge work, you are given an outcome and told to achieve it, the how is left to the knowledge worker (or AI). As AI is a probability machine the more detailed specification you give it, the more likely you are to get the result you desire, you are making the box for it to complete the task within.