6 min read
Artificial intelligence is prompting a new wave of predictions about software organizations.
AI may reduce the need for managers whose primary role is coordinating work. It has a very different effect on managers responsible for developing people. Before predicting the end of middle management, we should be clear about which kind of management we mean.
Teams may get smaller. Middle management may thin out. Leaders may spend more time working alongside AI agents, not just people. Some recent writing takes it further, suggesting organizations will need to reshape leadership roles and operating models entirely around this shift.
Before accepting those conclusions, I think there is a more important question to answer.
What do we actually mean by management?
If management is primarily assigning work, collecting status, coordinating activities, and monitoring execution, AI may significantly reduce the amount of that work. Intelligent agents are becoming increasingly capable of coordinating, generating, analyzing, and executing tasks that once required considerable management attention.
If management is fundamentally about developing people, the conversation changes.
In the organizations I led, managers up to the director level served as both individual contributors and people leaders. Their technical credibility mattered, as did the growth, engagement, and retention of the people they supported. Those human outcomes were the clearest measure of management success.
That experience taught me that technical contribution and people management can coexist without making task coordination the primary purpose of the manager. It also shapes how I think about the influence AI may have on organizational design.
Is This Really a New Management Model?
One emerging model describes the future manager as an outcome leader responsible for configuring value streams across people and autonomous agents.
That role requires systems thinking, product judgment, technical fluency, and accountability for results. Those capabilities matter, although much of the underlying responsibility is familiar. Strong technology leaders have long shaped teams, allocated capacity, connected strategy to execution, and remained accountable for outcomes.
AI adds a powerful new form of capability to that system. The more interesting question is whether coordinating autonomous agents and developing people should continue to be treated as the same management discipline.
AI will change software delivery dramatically. I am far less convinced that it requires a fundamentally new organizational structure.
For years, software organizations have been moving toward cross-functional, autonomous teams. Lean moved decisions closer to the work. Agile emphasized empowered teams. DevOps reduced barriers between development and operations. Product operating models encouraged long-lived teams accountable for customer outcomes. Team Topologies gave leaders better ways to organize around flow and cognitive load.
These ideas were already reshaping engineering organizations long before generative AI became mainstream. AI strengthens many of those decisions. It did not create them.
The Responsibilities Remain
The same distinction applies beyond management. Across the software delivery team, AI changes how responsibilities are performed more readily than it changes which responsibilities must exist.
Every successful software organization still needs to understand customer problems, prioritize investment, design solutions, build software, validate quality, protect systems, manage infrastructure, operate platforms, govern data, monitor production, and determine whether the work created meaningful business value.
Those responsibilities do not disappear because AI becomes part of the team. They become redistributed.
Historically, organizations, including the teams I led, distributed these responsibilities across specialists. Product managers focused on customer value, engineers built the software, quality engineers shaped testing strategy, and platform and security teams enabled reliable delivery. Engineering managers remained technically engaged while developing people and strengthening the organization around them.
AI introduces another capability into that system.
An AI agent may summarize customer research, analyze product usage, draft requirements, generate code, create tests, review pull requests, inspect telemetry, identify security risks, or produce documentation. These capabilities now extend across the entire software delivery team.
The execution may change, but the responsibilities remain, along with human ownership of the customer outcome, technical risk, and final accountability.
This is why I believe organizational design should begin with responsibilities rather than job titles.
When discussions begin with titles, the conversation quickly becomes which jobs AI will replace. Beginning with responsibilities leads to a different starting point: understand the work that must be done, identify the capabilities required to perform it, decide whether those capabilities should come from people, platforms, automation, or AI, and make ownership of the outcome explicit.
Smaller Teams Do Not Create a New Model
That perspective also explains why I believe AI may compress teams without inventing a new organizational model.
A cross-functional team that once required enough people to fill two pizzas may eventually operate successfully with one. Individuals may carry broader responsibilities. Platforms may provide more shared capabilities. AI may perform work that previously occupied multiple specialists.
Those are meaningful organizational changes. They are also choices organizations were already making before AI.
In organizations I led, Scrum Masters evolved into Agile Delivery Managers embedded within cross-functional teams. Later, through acquisition integration and broader team redesign, the dedicated role disappeared as delivery coordination, facilitation, and continuous improvement shifted to engineering managers and the teams themselves.
The structure changed, yet the responsibilities remained.
AI may enable similar redistribution across testing, documentation, analysis, operations, and coordination. The right design will still depend on the product, architecture, dependencies, risk, and team maturity. A smaller team with poorly assigned responsibilities remains poorly designed.
Managing Agents Is Different from Developing People
One experience continues to shape how I think about management.
Nearly a decade before generative AI became mainstream, our engineering managers and senior managers remained active technical contributors. Their management responsibilities were intentionally centered on people. They coached careers, provided performance feedback, developed future leaders, built engagement, supported retention, and strengthened the organization.
I wrote about this transition in 2022, describing management as a different profession built around trust, coaching, feedback, difficult conversations, and helping people grow. Those responsibilities shaped how we developed managers long before AI agents entered the discussion.
We deliberately kept people managers from becoming the daily authority over their direct reports’ delivery work. Whenever possible, managers and their direct reports served on different teams, preserving team autonomy while giving employees dedicated support for growth and performance.
Looking back, AI reinforces the value of those design decisions, even though they were made years before AI entered the organization.
A successful entrepreneur recently described one of the emerging leadership skills as learning to manage agents. I agree, although that responsibility is meaningfully different from managing people.
Agents require configuration, orchestration, governance, evaluation, and monitoring. People require coaching, mentoring, trust, difficult conversations, career development, and opportunities to grow.
Both matter. They are different responsibilities.
As AI performs more of the work, organizations may employ fewer people and need fewer people managers. The need for coaching, mentoring, engagement, performance reviews, and leadership development will scale with the number of people being supported.
The organizational model remains familiar, even as the balance between people and AI shifts.
Perhaps the more useful question is whether AI is separating two responsibilities that have historically been combined within one role.
Engineering managers have long balanced developing people with coordinating work. In command-and-control organizations, where management still centers on directing and monitoring execution, AI may reduce much of the role. In more autonomous organizations, the need for management remains tied to the number of people who require coaching, growth, engagement, and performance support.
That shift allows managers to build AI literacy while spending more time on the human responsibilities organizations have often struggled to prioritize.
Technology has consistently changed the economics of software delivery without rewriting the principles of good organizational design.
AI feels larger than previous technology shifts, and its impact may ultimately prove greater.
Even so, I suspect the organizations that succeed will be those that understand the work, distribute responsibilities intelligently across people, platforms, automation, and AI, and continue developing the next generation of human leaders.
What are you seeing in your organization?
Is AI creating a genuinely new management model, or is it finally separating people leadership from the coordination of work?
Further Reading
In The Shift to Software Management: A Skill Guide for New Managers, I explored the human skills required when a technical contributor becomes responsible for developing people, including trust, coaching, feedback, emotional intelligence, and career growth.
About the Author
Phil Clark is a technology executive, advisor, and author of Profitable Engineering (profitableengineering.com). His work focuses on helping leaders connect software delivery, operating discipline, and AI adoption to measurable business value.