Unknown AI involvement
Teams may know AI is used but cannot say where it materially produced the result.
ABOUT HELMARK
A standard for the gap between AI speed and human control.
Helmark answers one question: how do teams preserve human responsibility, verifiability and software understanding when an increasing share of software work is done by AI?
WHY IT EXISTS
AI changes how software is produced. Responsibility should remain explicit.
AI assistants and coding agents let a person accept large changes they did not create line by line. Existing SDLC practices still matter, but they do not always record how the result came about or whether the team can independently judge the proposed solution.
Helmark adds a narrow control layer: classify AI involvement, inherit default component criticality where mapped, scale verification, keep a human owner and make loss of understanding visible.
THE PROBLEM
“We use AI responsibly” is a statement. Helmark asks for evidence.
Uneven verification
A low-risk suggestion and an AI-assisted security change should not receive the same control.
Responsibility drift
“The AI did it” can become a substitute for conscious human acceptance.
Cognitive debt
Code can pass tests while the team's ability to explain and maintain it declines.
Helmark was started in 2023 as part of Scrumban Framework. It began as a response to a growing question: how should software teams use AI without losing responsibility, verifiability and a real understanding of the systems they build?
Over the following years, the topic developed through many discussions, meetings, panels and conversations with people involved in building and delivering software. Over time it became clear that the problem was broader than a single framework or way of working. That is why Helmark was separated and developed as an independent AI Delivery Governance standard.
We thank everyone who has contributed to its development so far — by sharing experience, challenging assumptions, pointing out problems and proposing alternatives. Helmark was not created only at a desk. It has been shaped by repeated discussion around a problem that continues to change.
We do not treat the current form of Helmark as final. We expect the standard to evolve together with AI, development tools and the way software is produced.
The direction of technology is increasingly clear: a growing part of software development may be performed autonomously by AI, and AI may perform more verification work as the technology matures. Under Helmark STD 2.0, however, AI cannot replace the human review required by the standard, and responsibility for acceptance remains human.
Helmark does not assume that we are already there.
Under Helmark STD 2.0, human responsibility for accepted software changes is required. AI can generate code, analyse solutions, write tests and support review, but teams are not always able to independently judge whether a proposed solution is correct, safe and aligned with the intended outcome.
That is the stage Helmark is designed for: how to use AI without losing control over what enters the software system.
We do not know exactly what software development will look like in five or ten years. We do know that a standard dealing with AI cannot remain static while AI itself changes continuously.
If you see something that should be changed, simplified, added or removed, write to us. Describe the problem, propose an alternative or share an experience from your own team.
We do not guarantee that every proposal will be accepted. We do guarantee, however, that every valuable proposal will be taken seriously.
Helmark will change together with AI. A standard for a changing technology must itself remain capable of change.
Send feedback