Agentic Engineering Can Change How the Public Sector Develops Digital Solutions

Public organizations are facing increasing digital pressure

Legislation must be translated into action faster. Citizens expect more connected digital services. Critical systems need to be modernized, while requirements for security, documentation, and operational reliability continue to rise.

Agentic engineering can become an important part of the answer. The concept describes a new way of developing software, where AI agents do more than support individual tasks. They become active participants in the development process. They can analyze requirements, propose solutions, produce code, prepare tests, review documentation, and identify dependencies. People still make the decisions, but AI takes on a larger share of the analytical and execution-oriented work along the way.

For the public sector, the most interesting aspect is not simply that development can move faster. It is that agentic engineering can create a more connected and traceable path from political intent and legislation to a finished digital solution.

From AI Tool to Digital Way of Working

Many organizations already have experience using AI as a support tool for text, documentation, analysis, or development. That is useful, but the impact is often local. One employee becomes faster. One team gets help with a defined task.

Agentic engineering is different. Here, AI becomes part of the delivery model itself. Multiple specialized agents work together to translate requirements into a solution, while people manage direction, quality, and accountability.

This requires a different discipline from traditional AI use. An agent does not just need a prompt. It needs access to the right context: requirements, architectural principles, security rules, legislation, data classification, technical standards, and previous decisions.

If that context is missing, AI fills in the gaps itself. That can produce proposals that seem convincing but do not fit the realities of the public sector. In a government context, that is a real risk.

In practice, this already looks different in organizations that are further ahead. At STAR, the Danish Agency for Labor Market and Recruitment, requirements workshops are recorded and transcribed. AI then cross-checks against relevant legislation, flags dependencies between requirements, and ensures internal consistency before human approval. Traceability from the legal basis to the requirements becomes part of the process.

Governance as Part of the Development Process

In private companies, speed can be a goal in itself. In the public sector, the picture is more complex. Digital solutions need to be efficient, but they also need to be explainable, auditable, and operated securely over time.

Agentic engineering should therefore not be seen as a shortcut around governance. It is an opportunity to embed governance more closely into the development process. AI agents can help ensure that requirements are not lost along the way, identify where a change affects other parts of the solution, compare tests with the original requirements, and support documentation as an integrated part of the work rather than as an afterthought.

That makes the technology particularly relevant for public organizations, where solutions must connect law, domain expertise, processes, data, and technology.

Human Direction Still Matters

Agentic engineering does not reduce the need for strong professional expertise. It changes where that expertise is applied.

Product owners need to formulate better requirements and assess whether AI’s proposals support the intended purpose. Architects need to define the boundaries for the technical choices agents are allowed to make. Legal and compliance specialists need to define the constraints the solution must operate within. Developers and testers increasingly take on the role of reviewers, quality assurers, and orchestrators.

This creates new demands for leadership. The organization needs clear principles for where AI is allowed to act independently, where human approval is required, and how decisions are documented. Agentic engineering requires more than technical maturity. It requires organizational maturity.

Suppliers Must Be Able to Show How AI Is Used

When agentic engineering becomes part of software development, it changes the requirements placed on suppliers. Public-sector customers should not simply ask whether a supplier uses AI. They should ask how: How is the quality of AI-generated output assured? How are decisions documented? How are security and compliance handled? Where is human control applied? And how are efficiency gains shared?

These will become central questions in future public-sector IT partnerships.

At the same time, agentic engineering changes the economics of digital development. If parts of the work can be carried out faster and in a more structured way, that should be reflected in collaboration models, contracts, and expectations for continuous improvement.

Start With a Concrete Development Problem

Public organizations do not need to begin with large transformations. A good first step is a defined area with clear friction: requirements work, documentation, testing, modernization of legacy systems, or quality assurance of changes. Here, agentic engineering can be tested in controlled settings where impact can be measured, and the organization can build experience.

But even small initiatives should be designed with governance from the beginning. Data, accountability, security, and documentation must be considered from the start if the experience is later to be scaled.

Agentic engineering can become a new digital discipline in the public sector. Not because AI agents should take over the development of mission-critical solutions, but because they can help create a more connected, efficient, and documented way of developing them. For public-sector decision-makers, the task is to find the balance between momentum and control, so the technology is used with the level of responsibility that public digitalization requires.