Author: Louise Fellows, Area Vice President for Sales, Northern Europe at GitLab
Agentic AI, the latest evolution of AI, lets developers focus on higher-level strategic initiatives, providing UK Government agencies with additional resources to support their missions.
Recent government plans underscore the potential of agentic AI for central government, prompting agencies to use AI to enhance value to taxpayers and drive operational efficiency. One area of immediate applicability is in agency software development.
Agentic AI enables a transformation by introducing intelligent systems that are capable of making independent decisions and executing complex tasks in collaboration with developers. This significantly accelerates the software development lifecycle, letting teams deliver software faster and maintain security and compliance standards. The capabilities are important for government, offering practical applications that automate failure remediation, code reviews, test generation, security policy enforcement, and change management.
Beyond increased efficiency, agentic AI supports cost-saving modernisation efforts, like reducing technical debt, fixing or patching security vulnerabilities, and refactoring legacy applications. Agentic AI could be the solution to decades-long challenges faced by government agencies, such as retiring outdated systems and legacy code like COBOL.
However, using AI agents effectively requires a shift in how we approach existing development processes, moving toward a human-AI collaborative environment. The new reality is not about eliminating the need for skilled software developers; instead, it’s about augmenting their capabilities and redefining their roles to address the public sector’s unique challenges.
To optimise the long-term benefits of agentic AI, agencies must rethink their development frameworks and processes. This begins with an iterative approach to AI implementation, establishing a foundational understanding of agentic AI, and eventually evolving into a continuous improvement cycle.
Step 1: Establish foundational comfort with AI assistance.
For those new to using AI, the first step is to build familiarity with and confidence in AI-assisted coding and documentation in low-risk areas. This ensures that agencies can build best practices and avoid pitfalls like data leaks and security vulnerabilities.
Once developers establish a basic level of comfort with AI, they can expand their use of AI in the software development lifecycle, specifically repetitive, time-intensive workflows where AI can create immediate value.
Step 2: Establish governance and interoperability standards
As teams become more comfortable using AI assistance in individual cases, agencies can start creating policies for AI tool use. These include data access permissions, security protocols, and quality standards.
The central government has encouraged agencies and departmental data leaders to establish criteria for data interoperability, to standardise data formats, and develop processes to address security risks. For government agencies that maintain confidential personal information, establishing standards lets agents operate inside the parameters of agency compliance and security.
These standards empower, rather than limit. Data protocols standardise how AI systems can share information and collaborate in platforms, and hhis ensures that agencies maintain data interoperability and consistent data formats.
Step 3: Strategically introduce and scale AI agents
Now, it’s time for the exciting part of the agentic AI journey; the deployment of AI agents to take on self-contained development tasks with some degree of autonomy.
Enabling autonomy expands the scope of tasks agents can handle, beyond the capabilities of a single agent and lets multiple agents collaborate on complex projects.
Agencies should refine developers’ skills for effective AI collaboration, like complex problem-solving and creativity. The partnership between AI agents and talented, AI-ready developers advances innovation and enables faster, more secure software delivery in support of government missions.
Step 4: Continuously improve through feedback and education
Agencies will need to iterate on their AI implementation workflows as AI agents achieve autonomous abilities. To ensure agencies maximise the value of their AI investment, it will be important to implement agent monitoring systems with clear metrics and correction protocols.
Ongoing education programmes for developers, IT leaders, and civil servants, especially those focusing on AI literacy, are important for continued benefits. Agencies can invest in AI literacy with training programmes that focus on prompt engineering, AI collaboration techniques, and system oversight. Working with AI is a complex skill that sets developers apart.
Embracing the future of AI-led development in government
Agentic AI is more than an incremental improvement; it redefines how we conceive, build, deploy, and maintain the software that powers the central government.
The transition to AI-led software development presents significant opportunities and strategic hurdles for public sector agencies. Agencies that embrace a shift as an opportunity not a threat will achieve the greatest efficiency and speed.
By addressing these key areas, from initial integration to governance and continuous learning, government organisations can thrive where AI capabilities augment human creativity and strategic thinking.
Author: Louise Fellows, Area Vice President for Sales, Northern Europe at GitLab