We are moving beyond process automation into the era of Agentic AI (autonomous systems that proactively achieve goals with limited human supervision). At the recent event, “Build to Scale: Secure AI in Action, From Governance to Execution,” Michael Kares, Head of Sales, and Adam Coard, Sr. Application Architect, from AOT Technologies, shared a roadmap for the future of government services. This future is defined by controlled intelligence within governed workflows.
The Shift: From Transformation 1.0 to 2.0
For years, “Transformation 1.0” focused on digitizing forms, creating portals, and basic automation. While these steps were necessary, they often left citizens struggling with complex government structures. Staff also faced challenges from manual data entry and switching between systems.
Transformation 2.0 represents a fundamental shift toward:
- Context Awareness: Systems that understand the “who, what, and where” of a request.
- Guided Journeys: Replacing long, static forms with AI-assisted submissions and dynamic data collection.
- Proactive Connectivity: Services that are connected and governed. They act with consent to pre-fill known data and reduce friction.
Orchestrate, Don’t Replace
AOT Technologies’ philosophy is based on a simple idea: “Don’t rip and replace, orchestrate.” Instead of tearing down effective legacy systems, the focus is on integrating them with modern, modular components.
| Key Enabler | Benefit |
| Common Components | Reusable building blocks that speed up government processes. |
| Model Context Protocol (MCP) | MCP is like a USB standard for AI tools. It lets models connect to different services consistently, without being tied to a specific vendor. It also introduces structured control over what systems the model can access and how. |
| Natural Language | This allows staff and citizens to interact with complex systems using everyday speech, rather than having to learn technical details. |
The “Hard Question”: AI in Regulated Environments
The challenge in highly regulated environments, such as government, healthcare, and financial services, is balancing AI’s creativity with the necessary safety. AI is inherently non-deterministic, whereas traditional software is deterministic. It is logic-based and reliable, but static. The solution is m8flow, a Python-based workflow engine that enables deterministic workflow execution.
“We can’t just let AI run havoc… There are things that have to happen in a very predictable fashion. So we still want to maintain deterministic rules.” — Michael Kares.
By embedding AI in controlled workflows, organizations ensure:
- Human-in-the-Loop: Critical decisions need human validation before submission.
- Audit Trails: Every action taken by an AI agent is recorded and can be explained.
- Zero Trust Security: Strict policies are followed while leveraging AI at every stage of the process.
Real-World Application: Employee Offboarding
To illustrate these principles, Adam Coard showed an automated offboarding process. Using MCP, AI agents were able to:
- Understand Policy: Read the PDF policy documents to determine the required steps.
- Gather Requirements: Identify the specific employee through a chat process.
- Execute Actions: Interact with tools like Jira and Clockify to deactivate accounts.
Once the AI finished its tasks, the process moved to a clear workflow where a human security team reviewed and approved the changes.
Watch the Full Presentation
Want to see these agentic workflows in action? You can watch the full presentation from the Build to Scale event below:
Conclusion
Whether it’s improving large public sector services or personal projects, like Adam’s work transcribing his grandfather’s WWII Spitfire logbooks with AI consensus and human review, the principle stays the same. By connecting the flexible potential of AI with the reliability of governed systems, we can create more value-added services that are efficient, transparent, and secure.
Does your organization struggle more with legacy system limitations or with managing the risks of new AI implementations? If so, contact us!