Inspiration

Public-sector decisions often suffer from fragmented workflows, inconsistent legal criteria, and inefficient manual review processes. These gaps create delays, increase legal risk, and reduce transparency.

POSE AI Agent was inspired by a simple question: what if an AI agent could help public institutions standardize decision workflows, detect risks early, and guide teams toward safer and more efficient outcomes?

The idea builds on the POSE framework: Process, Organization, Legal Security, and Efficiency. We adapted this concept into an AI-driven workflow assistant aligned with the GitLab AI Hackathon challenge.

What it does

POSE AI Agent is an AI-powered workflow assistant designed to support public-sector and compliance-oriented decision processes.

It helps teams:

  • Standardize administrative and decision-making workflows
  • Detect missing steps, inconsistencies, and legal/compliance risks
  • Recommend structured next actions based on predefined criteria
  • Improve traceability and transparency across public processes
  • Reduce friction in document review and approval flows

Instead of acting as a simple chatbot, POSE AI Agent is designed as an agentic workflow concept: it can react to process events, analyze structured inputs, and recommend or trigger next-step actions.

How we built it

We designed the project as an agent-oriented workflow concept compatible with the spirit of the GitLab Duo Agent Platform.

The project combines:

  • Structured workflow logic based on the POSE framework
  • AI-assisted analysis for compliance and risk detection
  • Rule-based recommendations for process standardization
  • Repository-based documentation and versioned project artifacts
  • A modular architecture that can evolve into custom GitLab agents and flows

The prototype and supporting materials were organized to demonstrate how this idea can be implemented in real administrative and compliance scenarios.

Challenges we ran into

One of the biggest challenges was translating a public-sector governance framework into a hackathon-ready AI agent concept within a limited time.

Another challenge was adapting a domain-specific solution (public decisions and legal process safety) to a developer-focused platform while still preserving the core value: turning governance, compliance, and legal safety into actionable workflows.

We also had to balance ambition with clarity, making sure the project remained realistic, understandable, and aligned with the hackathon requirements.

Accomplishments that we're proud of

We are proud of:

  • Turning a governance and legal-efficiency concept into an AI workflow project
  • Framing public-sector process standardization as an agentic automation opportunity
  • Creating a strong problem-to-solution narrative aligned with GitLab’s challenge
  • Building a submission that focuses on impact, compliance, and operational clarity

What we learned

This project reinforced that AI in software development is not just about code generation.

There is huge value in using AI agents to reduce friction in:

  • Compliance workflows
  • Approval pipelines
  • Process governance
  • Risk detection
  • Operational decision support

We also learned how important it is to design agents that do more than answer questions — they should observe workflows, reason about structured context, and help teams take action.

What's next for POSE AI Agent

Next steps include:

  • Turning the concept into a working custom GitLab agent or flow
  • Connecting workflow triggers to approval and review steps
  • Expanding rule sets for legal/compliance checks
  • Supporting document classification and decision templates
  • Building a more complete prototype for public institutions and regulated environments

POSE AI Agent can evolve into a practical solution for governments, public agencies, compliance teams, and organizations that need safer and more consistent decision workflows.

Built With

Share this project:

Updates