As artificial intelligence becomes increasingly capable of generating code, tests, documentation, and analysis, it is tempting to assume that many of the constraints shaping modern software delivery will disappear.
If machines can produce work faster than humans, why worry about frameworks, sprint planning, stand-ups, story points and all that?
But this assumption misses something fundamental.
The Constraint Was Never Coding Speed
The primary constraint in complex delivery systems has never been the speed of creating code.
The real constraints lie elsewhere:
-
coordination across teams
-
integration of components
-
decision-making at product and portfolio levels
-
validation of outcomes with customers
Even if AI dramatically accelerates the creation of software artifacts, these constraints remain.
In many cases, they become more visible.
Faster Creation, Slower Flow
AI increases the system’s ability to start work.
It does not automatically increase its ability to finish work.
This creates a dangerous dynamic.
As generation accelerates:
-
more features are started
-
more solutions are explored
-
more partially completed work enters the system
But downstream constraints—integration, validation, deployment—do not scale at the same rate.
The result is predictable:
-
work-in-progress expands
-
queues grow
-
delays increase
-
predictability declines
From the outside, activity increases.
Inside the system, flow deteriorates.
Moore’s Law may be dead—but Little’s Law is alive and running your delivery system.
The New Risk: Accelerated Overproduction
As the cost of starting work approaches zero, the temptation to start more work increases.
More initiatives.
More features.
More experiments.
Without constraints, the system fills with:
-
unfinished work
-
unvalidated ideas
-
partially integrated solutions
This is not acceleration.
It is overproduction.
And overproduction is one of the fastest ways to destroy flow.
Why Flow Control Matters More—Not Less
In an AI-assisted environment, flow control becomes more critical, not less.
Work-in-progress limits prevent the system from filling with half-completed work.
Aging policies expose items that are stalled in integration or decision queues.
Cycle time metrics reveal whether the system is actually delivering outcomes faster—or simply generating more activity.
The control loop remains essential because the system still needs a way to regulate how work enters and progresses.
“Agile Is Dead” — Or Misunderstood
The recent claim that “Agile is dead” reflects a narrow interpretation of Agile as a set of practices:
-
stand-ups
-
story points
-
sprint planning
-
specific frameworks
As technology evolves, some of these practices may change or fade.
But the underlying problem Agile attempted to solve has not disappeared.
Organizations still need to:
-
manage uncertainty
-
coordinate complex work
-
deliver outcomes predictably
If anything, AI increases the need for these capabilities.
The System Still Has to Answer One Question
Regardless of who—or what—is producing the work, the system must still answer a fundamental question:
How much work can we absorb while maintaining smooth, predictable flow?
Flow control provides the mechanism for answering that question.
One Model Still Applies
Across every level of the system—from portfolio strategy to team delivery—the same pattern holds.
Every delivery system can be understood as a simple control loop:
-
work flows through a value stream
-
signals such as WIP, aging, and cycle time reveal system health
-
policies define acceptable limits
-
when limits are exceeded, corrective action restores stability
Sensor. Controller. Actuator.
The same model applies everywhere.
A Final Thought
The tools around us will continue to evolve.
Frameworks will rise and fall.
Automation will increase.
AI will change how work is performed.
But the underlying challenge remains constant:
Systems must regulate how much work they attempt at once if they are to deliver reliably.
Regardless of whether the work is performed by humans or machines, the principle is the same:
Limit what you start. Maximize what you finish.