What AI Can, Can't and Shouldn't Do in NFP Roles - English

AI Realities for NFPs

New Zealand's not-for-profit sector entered 2026 at an inflection point. The government's AI Strategy, released in July 2025, explicitly extended voluntary responsible-use guidance to non-profits in a bid to accelerate adoption across the economy and deliver productivity gains. This policy push coincided with a sharp rise in practical AI integration: three-quarters of NFPs now deploy the technology for content creation, reporting and editing, while grant-writing usage climbed from 16% in 2024 to 33% last year.

The benefits arrive against chronic sector pressures. Many organisations operate with minimal staff and tight funding, making efficiency tools attractive for stretching resources further. Yet the concrete costs of missteps are rising. Privacy breaches involving client data can trigger regulatory action under the Privacy Act 2020, while inaccurate outputs risk eroding donor and community trust on which funding depends. Shadow AI compounds the problem: staff routinely use embedded features in everyday platforms such as Microsoft 365 or Canva without organisational oversight, creating compliance blind spots in organisations that lack dedicated IT or legal teams.

Non-obvious trade-offs complicate the picture. Larger NFPs adopt strategic approaches at roughly twice the rate of smaller ones, widening an internal digital divide. At the same time, the technology's limitations in grasping cultural nuance or ethical context sit uneasily with the human-centric nature of social services. Predictions for 2026 highlight industry-specific tools tailored to case management and community work, alongside the need for every professional to supervise AI outputs rather than simply use them. This shift promises to let tiny teams achieve previously unattainable scale, but it also demands updated job descriptions and training at a moment when only 14% of NFPs have any AI policy in place.

The Deloitte Australia episode in October 2025, where generative AI produced hallucinated citations and a fabricated judicial quotation in a major government report, offered a stark regional warning. The resulting partial refund on the AU$440,000 contract underscored how even sophisticated users can stumble, with consequences that smaller mission-driven entities can ill afford.