"Much of the context that makes large organizations tick is invisible to AI, such as political dynamics and historical decisions."
This might actually be a good thing. I think that the middle management layer (the "clay"-layer) is facing a lot of pressure for slowing things down. AI is empowering employees who want to get things done
I agree that AI is empowering employees who want to get things done at an individual or small team level, inside or outside of a large enterprise setting. I also agree that many enterprise companies had/have more middle managers than necessary, in part due to the pandemic hiring boom.
That said, for large organizations, the automation plot thickens when it comes to coordinating interdependent work. There are more relationships to manage and more stakeholders who shape decisions. I see two distinct branches here: (1) middle management for product teams (a layer that sits between execs and builders) and (2) "behind-the-scenes" functions that support product development, such as IT and HR. I'm more focused on the latter in this post. That's said, let's explore...
In both casses, humans still need to maintain situational awareness of what agentic systems are doing and why. The dynamic is like the shift from manual flying to autopilot. Automation handles execution, but pilots remain responsible for monitoring and interpreting context around them (e.g., weather conditions, something anomalous happening with the aircraft, etc). The pilot needs to intervene when conditions change.
When you pass a ticket from product to IT to procurement because someone is joining or leaving the company, it's like a multi-pilot chain. I'm skeptical that we can supply AI with all of that "context" (organizational, political, historical) so you could cut out entire layers.
For middle management: when working as intended, they provide structure, such as establishing process (e.g., structuring an evaluation programme). That's stuff that presumably neither execs or builders want to do, or if we were to somehow automate it, monitor.
All of this also assumes large enterprise organizations persist in roughly their current form. Maybe companies are no longer tens of thousands of people, but even if they shrink down to a fifth of the size, you still have complex organizations that need this human-context/AI orchestration layer.
Separate but related: the biggest risk for enterprise companies rn change management/organizational readiness. Going AI-native with a new team of five is very different from transforming a giant 20+ year-old company. The greater the organizational complexity, the harder the transition. I think that's why we have seen many failed AI pilots even though many ICs know the power of these tools.
That's a lot to unwrap right there! I think we need to be promoted to middle management to talk about it more! Just kidding, but yeah middle management does provide continuity which is a good thing
This gave me a good laugh, Kon! Yes, middle management *done well* (I swear it exists!) serves as the connective tissue that holds the moving parts together.
Agreed. Part of the challenge is that middle management working well is largely invisible, making it easy to underestimate its structural importance. The coordination chaos you shared is a great example of revealing what's being managed beneath the surface.
Something context you want a detox from:
"Much of the context that makes large organizations tick is invisible to AI, such as political dynamics and historical decisions."
This might actually be a good thing. I think that the middle management layer (the "clay"-layer) is facing a lot of pressure for slowing things down. AI is empowering employees who want to get things done
I agree that AI is empowering employees who want to get things done at an individual or small team level, inside or outside of a large enterprise setting. I also agree that many enterprise companies had/have more middle managers than necessary, in part due to the pandemic hiring boom.
That said, for large organizations, the automation plot thickens when it comes to coordinating interdependent work. There are more relationships to manage and more stakeholders who shape decisions. I see two distinct branches here: (1) middle management for product teams (a layer that sits between execs and builders) and (2) "behind-the-scenes" functions that support product development, such as IT and HR. I'm more focused on the latter in this post. That's said, let's explore...
In both casses, humans still need to maintain situational awareness of what agentic systems are doing and why. The dynamic is like the shift from manual flying to autopilot. Automation handles execution, but pilots remain responsible for monitoring and interpreting context around them (e.g., weather conditions, something anomalous happening with the aircraft, etc). The pilot needs to intervene when conditions change.
When you pass a ticket from product to IT to procurement because someone is joining or leaving the company, it's like a multi-pilot chain. I'm skeptical that we can supply AI with all of that "context" (organizational, political, historical) so you could cut out entire layers.
For middle management: when working as intended, they provide structure, such as establishing process (e.g., structuring an evaluation programme). That's stuff that presumably neither execs or builders want to do, or if we were to somehow automate it, monitor.
All of this also assumes large enterprise organizations persist in roughly their current form. Maybe companies are no longer tens of thousands of people, but even if they shrink down to a fifth of the size, you still have complex organizations that need this human-context/AI orchestration layer.
Separate but related: the biggest risk for enterprise companies rn change management/organizational readiness. Going AI-native with a new team of five is very different from transforming a giant 20+ year-old company. The greater the organizational complexity, the harder the transition. I think that's why we have seen many failed AI pilots even though many ICs know the power of these tools.
That's a lot to unwrap right there! I think we need to be promoted to middle management to talk about it more! Just kidding, but yeah middle management does provide continuity which is a good thing
This gave me a good laugh, Kon! Yes, middle management *done well* (I swear it exists!) serves as the connective tissue that holds the moving parts together.
Agreed. Part of the challenge is that middle management working well is largely invisible, making it easy to underestimate its structural importance. The coordination chaos you shared is a great example of revealing what's being managed beneath the surface.