Can We Really Remove All the Managers?
The latest theme rippling through Silicon Valley and organizations is a belief that huge efficiencies and cost savings can be captured by removing layers of “middle managers”. An oversimplified description of this trend goes something like this:
Remove all the people working between executive management (i.e. the C-level folks) and individual contributors (ICs)
Fill the vacuum created by displacing managers with AI tools, platforms, and agents so the remaining C-level and IC folks all become instantly hyper-productive
Iterate on #1 and #2 above to magically scale revenue and operating profits
Make all stakeholders insanely wealthy, well, at least those who hold preferred equity and founder common shares
Realities of the Player-Coach
Embedded in the scenario above is the concept that everyone — from the CEO to the most junior-level IC — needs to become a “Player Coach”. Ah, the PC. That unicorn who can simultaneously lead, manage, be strategic, and do all the detailed, daily tasks over and over and over again that ultimately moves a company forward.
Keep in mind, Phil Jackson is arguably one of the top 3 NBA coaches of all time, but he was a pretty average (maybe even below-average) NBA player. Ted Williams may be the greatest hitter in MLB history, and was also one of the worst managers/coaches in MLB history.
The concept of “Player Coach” has been around forever, and it isn’t new to Silicon Valley either. I can’t personally recall a time as a manager or even as a startup CEO that I wasn’t also doing “IC” type work within one or more functional areas. That said, my IC work was always suboptimal compared to our IC experts, but I added it to my to-do list. It was how we got more done and could “play bigger” as a company.
This latest AI-driven push towards the utopia of the player-coach means one thing for sure. The workloads will scale dramatically for the humans drafted into these PC roles. Now I’m biased, of course. I’m a human, but I imagine that the pendulum will swing back to recognize that human managers will outperform a world where organizations subsist on a diet of PCs alone.
The Case for Keeping Human Managers
Why do I think human managers will outperform a world of AI-powered PCs:
Humans will always have marginally more detailed and nuanced context for understanding an organization’s goals within the broader strategic context of competitive threats, partnerships, and balancing opportunities and risks. So long as we don’t abdicate this advantage to AI, I imagine AI will always be saying “Nice catch!” to us.
A corollary to the bullet above, humans will excel at planning over long horizons while AI will excel at getting short-term tasks done. Given our incremental context advantage, human managers will be better at looking around corners, asking the “what if” questions, and appropriately challenging the assumptions from the C-suite and the work product of ICs.
Managers make decisions, over and over. AI builds scenarios and then demurs, always essentially saying “well, you could do this or do that, and here are the pros and cons.” And C-level leaders relying solely on the work product generated via AI by ICs will lack the critical “decision-making” scrutiny professional managers are most skilled to provide.
Responsibility and accountability go hand in hand, and they are embedded in the fabric of how humans operate in a company. Somewhere along the line “the buck stops here”. Ultimately we can point to the CEO, but there are many decisions that are connected to responsible and accountable operators before the CEO, and these are the managers. When AI makes a mistake or just fails to perform, how will it be held responsible and accountable for its actions?
The multidimensional ability to guide, coach, support, and conduct difficult conversations with other humans is embedded in the best managers. How will these characteristics evolve in AI managers who today are unidimensional and oftentimes just sycophantic - as well as wrong and incomplete - in their responses. Can we expect AI managers to ever feel any shame in being wrong and take any meaningful stake in being right?
No doubt people are going to be asked to do more and increase their personal span of control. They are going to be asked to carry the entirety of the organization in their head, or at least be able to continually prompt Claude to play it back to them.
This will be cognitively exhausting, and will likely lead to not just human burnout for those tasked with trying to be “a LeBron James” in their organization, but it will also be prone to errors and costly missteps. Maybe these hiccups won’t matter. Maybe the pace of iteration and “fixing bugs” will be so fast with AI that the costs associated with losing the managers will be justified.
I remember when holacracy was all the rage in Silicon Valley. The spectre that companies could remove the hierarchical org chart and replace it with flowing circles of self-governing teams that somehow always knew what to prioritize and how everything aligned with the company-level success metrics. I’m pretty sure holacracy hasn’t become anything close to the norm in Silicon Valley, let alone anywhere else.
Sure, the “what’s different now?” answer may very well be that AI changes the game so dramatically that the technology will fundamentally change how organizations are designed. And yes, again, I’m a biased human making the case for humans, and in particular humans called “managers” (all of whom are already PCs by the way), to continue being relevant and valuable to organization’s design.
So how does this play out? I don’t know. Let’s see how organizational design evolves and how the work humans are doing changes over the next year. As is always the case, time will tell so long as we really pay attention.
NextPlay>Forward AI Disclaimer: I very actively use artificial intelligence and large language models to generate the content you read here, but I do review it and edit it to make sure it can be generally useful to people who read it. Keep in mind that AI can make mistakes - check important information. Let me know if I make any errors and I will correct them.