TrIAje: AI-Assisted Legal Guidance Tool

The problem

Civil society organisations (CSOs) often face complex legal obligations while operating with limited access to specialised legal support. In Peru, regulatory requirements related to governance, international cooperation, tax registration, data protection, and organisational formalisation can be difficult for smaller organisations to interpret and act on.

For many CSOs — particularly smaller or emerging organisations — this creates serious operational challenges. They must navigate greater uncertainty around compliance, governance, and reporting obligations in a context where mistakes or ambiguity can affect their ability to operate, access resources, or collaborate with others.

At the same time, professional legal assistance in Peru is expensive and typically charged by the hour. For organisations with limited budgets, spending time in a meeting while a lawyer tries to understand what the organisation does and what it needs can consume a significant portion of the available consultation time. Many CSOs simply do not have fast, affordable access to the specialised legal guidance they need.

Most available legal technology tools are designed for legal professionals, rely on subscription models, and use technical language that can be difficult for non-specialists. The team identified a gap in accessible, free tools designed specifically for civil society organisations navigating these regulatory requirements.

The combination of rising regulatory pressure and limited legal capacity creates a structural vulnerability for civic action in Peru.

The solution

TrIAje is a free, open-source Artificial Intelligence (AI)-assisted legal guidance tool designed to help CSOs in Peru navigate the country's regulatory environment. It takes its name from IA, the Spanish acronym for artificial intelligence, and triage, the process of guiding people to the right kind of support. The tool provides orientation on issues such as how to formally establish an organisation, governance requirements, compliance obligations, tax registration, data protection, intellectual property, non-disclosure agreements, and collaboration frameworks between organisations.

The tool prioritised source-grounded AI over general-purpose generation, recognising that legal orientation requires traceability, accuracy, and user verification. That is why it is built on a Retrieval-Augmented Generation (RAG) approach — meaning it does not generate answers from general knowledge, but draws from a curated body of legal information specific to Peruvian civil and administrative law. This knowledge library was developed in collaboration with a legal firm with prior experience in legal technology and covers approximately 10 topic areas identified through interviews with 15 civil society organisations regarding their most common legal pain points.

How it works
True to its name, TrIAje works as a first point of orientation through a straightforward chat interface, similar to other AI assistants. For example, when a user asks, “How do I open a non-profit that works with youth?”, the tool responds with a structured answer including the relevant legal framework, the requirements that need to be met, a step-by-step process for registration, applicable tax obligations, and key considerations to be aware of. It also provides links to source material so users can verify information further.

The tool generates practical examples based on the user's question. If someone asks about setting up a youth sports organisation, TrIAje will walk through a concrete scenario — three people deciding to formalise a collective, drafting their founding documents, registering with the relevant authorities, and obtaining tax identification. Where the question involves working with vulnerable groups such as young people, the tool automatically flags additional legal protections that the organisation should be aware of.

Users can also upload their project descriptions or initiative plans, and TrIAje will provide legal feedback specific to that project — including what organisational structure might be appropriate and what regulatory steps would be involved.

When TrIAje encounters something an organisation cannot do under its current legal structure, it does not simply say no. Instead, it suggests alternative approaches and provides the steps to do so.

The tool also prepares a “legal package” — a set of structured questions and an introduction to the organisation's work designed to be taken into a meeting with a lawyer, so that paid consultations can focus on the specific legal questions rather than background explanation.

TrIAje does not replace lawyers. It helps civil society organisations understand the legal dimensions of their work, identify the questions they need to ask, and use scarce legal consultation time more effectively.

Security and data protection
Because CSOs may input sensitive information about their work, the tool was built with a strong emphasis on data security. It uses an end-to-end encrypted AI model service that ensures data privacy and secure processing throughout the workflow — meaning that data entered into the tool is encrypted, so neither the AI provider nor any third party can access it. The team chose not to use mainstream Large Language Model (LLM) providers specifically because those platforms typically store user inputs. The tool is registered with Peru's national data protection authority and has published terms and conditions and a privacy policy.

TrIAje also includes clear disclaimers defining its role as a first-layer support that helps organisations reduce uncertainty and prepare for more effective engagement with legal professionals.

The impact

Early use of the tool showed demand for practical, context-specific legal orientation. The tool was first introduced in a capacity-building setting with 31 civil society organisations from 10 different regions across Peru, and it was user-tested with a smaller group of organisations who provided feedback that informed the final iteration.

During the boot camp, approximately one-third of participants actively used TrIAje while developing impact models for their organisations. Organisations used the tool to answer immediate practical questions — for example, checking whether their organisational type permitted them to sell research or consultancy services to the private sector, and understanding what conditions applied.

Because the tool is in an early stage of implementation, formal impact data is still being collected. The team plans to continue gathering feedback from the 31 organisations as they use the tool in their day-to-day operations.

Lessons learned

• Reliable legal AI depends first on a reliable legal knowledge infrastructure. Laws are subject to amendments, and the current, accurate versions are not always straightforward to find through general web searches. Only legal professionals tend to know where to locate the most up-to-date sources. Building a reliable knowledge library required partnering with a legal firm — and specifically one that understood legal technology.
• AI tools built on external infrastructure require ongoing monitoring and fallback plans. During development, the language model provider changed the underlying model without notifying the team, causing the tool to stop working. The issue was resolved within a couple of hours, but it was not something the team had anticipated. Any team building on third-party AI services should incorporate regular checks to ensure the tool continues to function as expected.
• Adoption improves when digital tools are embedded into real decision-making moments. The team embedded TrIAje into the boot camp programme at the point where organisations were making decisions about their initiatives and needed legal clarity. Introducing the tool when it was immediately useful — rather than as a standalone demonstration — helped drive adoption.

The project was shaped by Peru’s regulatory framework for civil society organisations, particularly rules related to organisational formalisation, governance, tax registration, data protection, intellectual property, and international cooperation.

A key contextual factor was an amendment to the law governing the national Agency for International Cooperation (APCI), which expanded state oversight over organisations receiving international cooperation funds. Civil society and human rights groups have raised concerns that this change could enable excessive control, arbitrary restrictions, and potential censorship of civic activity. The regulatory environment also carries significant consequences for non-compliance, including fines that can reach up to half a million US dollars — an amount far beyond the annual resources of many small organisations.

These conditions shaped the design of TrIAje in three ways. First, the tool was built around a curated legal knowledge base grounded in Peruvian civil and administrative law, rather than relying on general AI-generated responses. Second, it was designed to help organisations identify relevant obligations and prepare more focused questions for professional legal advice. Third, because users may input sensitive information about their work, the tool incorporated privacy safeguards, published terms and conditions, and registered with Peru’s national data protection authority.