AI Is Accelerating the Data Plane. Who’s Building the Control Plane?
Every week brings more news about how AI is transforming software development. AI writes code, generates tests, reviews pull requests, creates documentation, explains unfamiliar APIs. The productivity gains are real. Yet many organizations still struggle to deliver value predictably.
Why?
One reason is because most AI investment today is focused on accelerating the data plane, while the control plane remains largely unchanged.
The Two Planes of Enterprise Delivery
Modern distributed systems distinguish between the data plane and the control plane. The data plane performs the work. The control plane regulates how the work flows.
The same distinction applies surprisingly well to enterprise software delivery. Today’s AI tools are deployed primarily on the data plane. They accelerate execution, but they do not yet regulate the delivery system.
| Data Plane | Control Plane |
|---|---|
| Write code | Decide what to build |
| Generate tests | Limit work in progress |
| Create documentation | Prioritize work |
| Refactor code | Detect aging work |
| Review pull requests | Trigger management intervention |
| Answer technical questions | Forecast delivery |
| Create designs | Balance demand against capacity |
Data Plane
The data plane is where value is created.
It includes activities such as:
- Discovery
- Development
- Testing
- Deployment
- Release
This is where today’s AI tools excel.
GitHub Copilot, Cursor, ChatGPT, Claude, Gemini, and similar tools help teams execute work faster and with less effort.
They are exceptional data-plane accelerators.
Control Plane
The control plane governs how work moves through the delivery system.
It answers questions such as:
- What should we build next?
- Are we exceeding our WIP limits?
- Which features are aging?
- Where are the bottlenecks?
- Are we delivering predictably?
- What intervention should leaders make?
These are fundamentally different problems. They are not about writing software. They are about regulating a complex system.
Faster Coding Doesn’t Guarantee Faster Delivery
Suppose AI doubles a team’s coding capacity. Does delivery double? Not necessarily.
The bottleneck often moves elsewhere:
- Poor feature definition
- Excessive work in progress
- Decision latency
- Cross-team dependencies
- Approval queues
- Overloaded backlogs
The organization has accelerated execution without improving how work is managed.
The result is familiar. More code. The same delivery problems.
The Next Competitive Advantage
What AI is doing today is largely local optimization:
- write code faster
- generate tests faster
- review PRs faster
What needs to happen next is system optimization:
- reduce decision latency
- reduce WIP
- improve prioritization
- reduce dependencies
- improve predictability
Today’s AI optimizes individual tasks. Tomorrow’s AI must optimize the entire value delivery system.
Imagine an AI assistant that continuously:
- detects WIP limit violations
- identifies aging work before it becomes a delivery risk
- forecasts completion dates from actual throughput
- recommends re-slicing oversized features
- prepares Flow Review agendas
- highlights emerging bottlenecks
- suggests where leadership intervention will have the greatest impact
Notice what has changed. The AI is no longer performing the work. It is helping regulate the system that performs the work.
Looking Beyond AI Coding Assistants
If the objective is to deliver customer value faster, then coding assistants are only the beginning. As software development becomes increasingly automated, competitive advantage will depend less on how quickly organizations can write software and more on how effectively they regulate the flow of work from strategy to delivery.
The organizations that thrive won’t simply have faster developers. They’ll have better control systems.
The next frontier in enterprise agility isn’t accelerating the data plane it will be to finally build the control plane.