The product
OpenCity AI Assistant provides all institution staff with a single AI-powered work environment. Unlike generic tools, it operates in the specific context of Italian Public Administrations: it knows the relevant regulations, administrative processes, institutional language and public service standards.
The institution does not need to adapt a tool designed for others: it has a ready-to-use environment, aligned with the way public administration works, from day one.
Interaction happens in natural language: simply describe the problem or task at hand. The AI operates in three fundamental modes: it suggests and drafts responses, content and communications that the officer can review at any time; it monitors and alerts about regulatory deadlines, supplier SLAs and missing publications before they become problems; it analyses and reasons on aggregated data, reports and performance indicators to provide a strategic overview.
Every action remains under the officer’s control, following the human-in-the-loop principle: the AI proposes, suggests and accelerates, but the final decision is always human.
The three assistants
The work environment is structured around three assistants, each specialised in a specific operational area of the institution.
Staff Assistant is the daily reference point for every employee. It assists with writing texts, organising deadlines and searching for acts and regulations. It directs users to the right service: IT support, HR, training, maintenance.
Tech Assistant is the right-hand tool for the IT office and the Digital Transition Manager. It analyses and compares supplier solutions on interoperability and security, manages technical reports via integrated ticketing and supports planning the institution’s three-year IT plan.
Governance Assistant is the compass for leadership. It queries reports and data in natural language to evaluate scenarios, monitors Transparency and operational plan progress, verifies act compliance and produces draft opinions.
How it works
At the core of the three assistants is an open source AI generative product distributed through OpenCity Italia’s ACN-qualified SaaS cloud infrastructure.
The technological core is based on an Agentic architecture with RAG (Retrieval-Augmented Generation) and Agentic RAG capabilities — the same approach that powers the OpenCity chatbot. Unlike systems based on generic AI, this solution anchors every response to verified data from the start. The solution is model-agnostic and adopts the MCP protocol to ensure interoperability with other institutional or third-party solutions.
Each assistant provides the officer with four fundamental tools: the chat for natural language conversation and content generation; tasks for organising work into traceable assignments with deadlines; notifications with which the assistant proactively alerts about deadlines, anomalies or content to update; the dashboard for an instant summary view of key indicators.
To learn more, read the complete documentation.
Knowledge base
To operate in the specific context of each institution, the assistants draw on a pre-loaded knowledge base that grows over time without complex technical intervention.
From day one the assistant consults national sources: regulations, AgID guidelines, enabling platforms and the Three-Year Plan. The institution can then add regulations, circulars, resolutions and spreadsheets for increasingly precise and contextualised responses.
Behaviour is guided by pre-loaded instructions: content templates, editorial rules and formats for acts. These are also customisable: the institution can define approval workflows, ticket steps and specific communication styles.
The assistant does not just find information: it knows how to use it, and becomes increasingly useful as the institution feeds it with its own documents and rules.
Benefits
Thanks to this architecture and specialised knowledge, OpenCity AI Assistant offers concrete advantages over both working without AI and using generic tools:
- Immediate operational responses: the assistant provides precise guidance on procedures, regulations and institutional content without the need to search through documents and circulars
- Content production support: from drafting acts to composing communications, the assistant proposes drafts that the officer can accept, modify or reject
- Quality monitoring and control: the assistants flag missing content, approaching deadlines and inconsistencies in published materials
- All information in one place: the assistant draws on national regulations, local rules and the institution’s website content, reducing source fragmentation
- Specialised knowledge: unlike ChatGPT, Gemini or Deepseek, the assistants operate in the context of the Italian PA and the institution’s specific information, not generic knowledge. The assistant also adapts the focus of responses based on the user’s role within the institution.
- Fewer hallucinations: when the assistant lacks reliable information, it says so and invites verification with competent colleagues, rather than inventing plausible answers
- Traceability and control: every proposed action is visible and approvable by the officer; the system maintains a log of all interactions, in compliance with AI regulations for public administration
Transparency and privacy
Efficiency benefits do not come at the cost of security. OpenCity AI Assistant is designed so that the institution does not need to worry about compliance, security and infrastructure: these are handled upstream by OpenCity Labs.
The solution operates on ACN-qualified infrastructure; the LLM used is Gemini via Google Agentic Platform, with a model-independent architecture ready to evolve without lock-in. On privacy, a control filter is active: if personal citizen data is being sent, the system warns the user. If the user proceeds, data is automatically anonymised before transmission. Every interaction is logged (who, when, outcome), exportable as CSV and compliant with AgID Guidelines principle P.14, the EU AI Act and Law 132/2025. The institution retains full control through granular permissions: it decides who can access the assistants, upload sources and consult system logs.
Like any technology, AI can make mistakes: in line with the technical principles for AI use, the user is informed of this possibility and invited, in case of doubt, to verify information through ordinary channels.
Regulatory compliance
Transparency and privacy are part of a broader framework. OpenCity AI Assistant is designed in compliance with European and national standards for public sector transformation:
- Security, guaranteed by ISO 27001 certification and ACN qualification of OpenCity Italia
- Privacy, strict compliance with GDPR and a control filter for any citizens’ personal data
- Open source. The preferred choice according to technical rules for software procurement in public administration
- Interoperability. Adoption of open standards (MCP)
- User experience. Design and accessibility guidelines
- Technical rules on AI in public administration. Compliance with Law 132/2025 and the EU AI Act