The Modernisation Control Papers — 01
AI has made code cheaper. It has not made change safe.
AI is reducing the cost of producing software change. The scarce capability is increasingly the evidence, authority and control required to decide which changes are justified.
Enterprise modernisation is approaching an inversion point.
For decades, producing change was expensive. Teams spent months understanding legacy systems before altering them.
Autonomous agents are changing that. They can inspect large estates, recover business rules, map dependencies, generate specifications, propose architectures, transform code and create tests at increasing speed.
The constraint has moved.
It is no longer the production of change. It is deciding which changes are justified. What evidence supports them. Which assumptions remain unresolved. Who accepts the consequences. Whether the eventual outcome still matches the original intent.
More generated code does not remove uncertainty. It can multiply it.
A confident answer without provenance is not evidence. A completed agent run is not an approved decision. A passing test is not proof that the right system was built.
Modernisation therefore needs a control system connecting six things:
- Evidence — what was actually observed in the estate.
- Interpretation — what people and agents infer from it.
- Decision — what the organisation chooses to do.
- Approval — who accepts the risk and consequence.
- Execution — what is changed, by whom, and under which constraints.
- Outcome — whether value was delivered without losing required behaviour.
The distinction matters. Evidence must remain distinguishable from inference.
Assumptions must remain visible rather than hardening silently into architecture. Material decisions need named owners. Generated artefacts need provenance. Exceptions and second opinions need durable records. Outcomes need to be tested against the decision that authorised the work, not merely against the code that happened to be produced.
This is not a case against autonomy. Agents should investigate, propose and execute wherever the evidence and risk permit.
Human intervention should occur at consequence boundaries, not as ceremonial approval after the work is already done.
The operating principle is simple:
Increase autonomy as evidence strengthens. Increase scrutiny as consequences grow.
The emerging market for AI-augmented code modernisation shows how quickly engineering capability is advancing.
Gartner's latest research describes a shift towards tool-led, continuous modernisation, with governance, auditability and human oversight becoming important enterprise considerations.
Harten's position is that these are not supporting features. They are the architecture of safe change.
We call that discipline Modernisation Control: the evidence and decision layer governing human-and-agent transformation.
Start with evidence. Reduce uncertainty before introducing change.
Originally published by Harten Technologies on LinkedIn. This harten.io page is the canonical archive.
Where this becomes operational
Apply the thinking to a real application.
If the problem described here exists in one of your applications, Harten can establish the current evidence, unresolved uncertainty and the basis for the next decision.