The Future of NGO Technology

Agentic AI Checks In

An important project report is due. Lily, an NGO associate, needs to gather information from databases, CRMs, financial platforms, and document repositories. She moves between these systems, searches through documents, identifies the relevant information, pulls it together, drafts the report, and prepares it for submission to the donor.

Now imagine if technology could do much of this work for her.

What if, instead of doing each step herself, Lily could hand the process to an AI system that could gather the information, identify information gaps or inconsistencies, draft the report, and bring that report to her for review?

That is the next stage of NGO technology: agentic AI systems that can take appropriate action and keep the process moving forward.

But because the system can access information and take actions, it also needs clear boundaries around what it can access, what actions it can take, and when a person needs to step in to ensure that no risky actions are taken.

What Agentic AI Needs

For this kind of AI to operate safely and responsibly within an organisation, certain boundaries need to be built into the system itself. Five elements are particularly important:

• Access Controls: Define what information, systems, and tools AI can access and what actions it is permitted to take. Limiting access to what is necessary reduces the potential impact of errors.
• Organisational Memory: Give AI access to current, approved organisational knowledge, including policies, project documentation, reports, and relevant requirements, so it has the context needed to carry out work appropriately.
• Human Oversight: Establish points where people review, approve, or intervene when a task requires human judgement or an action has significant consequences.
• Exception Handling: Define how the AI should respond when information is missing, conflicting, or outside its defined scope, ensuring that it does not fabricate information or proceed without sufficient information.
• Auditability: Record AI activity so the organisation can see what actions were taken, review decisions, identify problems, and maintain accountability.

The goal of good governance is not to limit what agentic AI can do, but to give it enough autonomy to move work forward while ensuring people remain in control where it truly matters.

What Agentic AI Could Mean for NGOs

By becoming part of how work is carried out, a governed agentic AI system could fundamentally change how an organisation operates, building greater capacity, supporting continuity, and enabling smarter decision-making:

• Greater capacity: By taking on routine, repetitive processes, AI can reduce the time staff spend on administrative and coordination work and free them to focus on decisions, relationships, and mission-critical work.
• Continuity: With organisational memory, AI can continue to access important organisational knowledge even as people move between projects or leave the organisation. This helps ensure that changes in personnel do not disrupt the AI's ability to carry out its work.
• Smarter decisions: By accessing and bringing together information from across the organisation, AI can potentially surface patterns, gaps, and opportunities that might otherwise go unnoticed, helping staff make more informed decisions.

Conclusion

For Lily, the difference is simple. Instead of spending her time moving between systems, searching for information, and coordinating each step of a process, she can exercise her judgement where it matters and use the time she gains to focus on building relationships, supporting her team, solving problems, and advancing the organisation's mission.

That is the opportunity of governed agentic AI: technology can take on more of the work while people spend more of their time on the work that requires their expertise, judgement, and human connection.