Most discussions of the impact of AI on the world of work, both optimistic and pessimistic, focus on its impact on what I would describe as front-line jobs or as jobs whose output is tangible. This includes call center workers whose jobs might be automated away, as well as people writing code, writing copy, or creating images, who may either also be automated away or made vastly more productive, depending on who you ask.
Yet if the AI revolution is in any way similar to the Industrial Revolution, it will not only change, create, and eliminate specific jobs but change the way we organize work. During the Industrial Revolution, not only did people start using heavy machinery a lot more, but both industry and, eventually, agriculture became vastly more concentrated. While tractors and big machinery in the factories were much more productive than artisans and smallholder farmers using animal labor, they were simply not economical to run except at scale. So the technology change drove the change in what work meant at the scale of the whole society.
Embracing the potential of AI will lead to a similarly deep change. Consider the many jobs that are not on the front line. I am referring to middle and senior management, who form a key part (both in terms of decision-making and staff numbers) of any medium or large firm. If you’ve ever worked in such a firm, I’m sure you could name some managers who did little more than consume natural-language input (emails, reports, conversations) and produce natural-language output according to some pretty simple hard-wired script in their head. Of course, a good manager does far more than that – but in any case, rule-based processing of natural-language information (compiling reports, etc) is an important part of a management job.
The reason this has never been automated is because until recently, computers in general and machine learning techniques in particular were really bad at working with natural language. Now that this has changed, we can expect that kind of automation to appear, redefining what a management job is about. To make the most of this change, the organizational structure of the firm would also most likely change, in the same way that a factory was not just a very large artisan’s workshop.
What might that look like? Consider the firm as a system that aims to coordinate those working in it toward a common task. Two key roles of middle management in a classic firm are, firstly, to aggregate and feed up information about the state of affairs on the ground; and, secondly, to flesh out and implement the decisions conveyed to them by their leads.
With the present state of AI, it’s already possible to imagine that the first one of these functions can be almost or entirely automated. So, a CEO or a manager at any level of the firm could ask a question in natural language about anything going on in the firm (as long as it leaves an electronic trace) and get an aggregated and understandable answer in natural language, including any relevant numbers. Would that merely free up middle management to focus on other aspects of their role, such as people management and communicating the organizational intent to their direct reports – or would that enable an entirely different kind of firm to thrive and scale?
I think it’s the latter. The fundamental concept of that transformation is autonomy. When I first joined Wise (then Transferwise) Treasury four years ago, I was amazed at how far it took the culture of autonomy; both at single-person and team level, there was freedom to decide what matters in your domain, and to work on that; and the role of the lead was not to tell their reports what to do, but to enable them to do what they thought needed doing. The effectiveness and the human flourishing unleashed by this approach were a revelation to me.
Yet while autonomy-centric organizations are great to work at and great at local optimization, they can be a challenge to scale. If each team is free to choose, within its domain, what to work on, how can the next level of leadership ensure all critical areas are covered, and there is not too much duplication of effort? This is mostly an information challenge – if gaps or duplicate efforts are pointed out to the relevant autonomous teams, they will be happy to address these; the challenge is spotting them.
And this is exactly what AI can help with. If every person working at the firm spends at least a little time writing down what they do and why (surely not too much to ask), AI could aggregate that to the level and perspective needed and thus allow coordination without the need for top-down control.
This change will not be easy – too many standards, such as audit requirements for listed companies, enforce a top-down, cogs-in-a-machine structure of companies. However, if AI, by taking information exchange and coordination to the next level, enables firms to combine the innovation and human-friendliness of autonomy-centric structure with the advantages of scale, such firms will naturally thrive and lead by example and, in time, become the new norm.
Thus, while the Industrial Revolution was, in many ways, dehumanizing to the workers, the information exchange enabled by this generation of AI has the potential to promote human flourishing at scale by supercharging autonomy-centric firms.
