Le jour où l’agent conversationnel est arrivé : les OBNL et l’IA générative, trois ans plus tard

Some technologies arrive gradually; others seem to arrive overnight. On 30 November 2022, OpenAI released ChatGPT to the public, and within months, generative artificial intelligence had moved from research laboratories into the daily routines of millions — including the staff and volunteers of countless nonprofits.

Having spent many years in technology leadership, I have witnessed several such waves. What distinguishes this one is not only its speed, but the intimacy of its reach. It touches writing, reasoning and communication — the core activities of mission-driven work.

The promise, honestly stated

For small organisations, the promise is real. A tool that helps draft a grant narrative, summarise a lengthy report or translate a notice into several languages can return precious hours to people who never have enough of them. Used thoughtfully, it can narrow the gap between large, well-staffed organisations and small ones.

The cautions, equally honest

Yet the same tools demand discernment. They can produce confident errors. They can reproduce biases. And they invite a subtle temptation: to share information that should never leave the organisation — a client’s circumstances, a donor’s details — in pursuit of convenience.

These are not reasons to refuse the technology. They are reasons to govern it.

Principles I would commend

• Keep people accountable. Every AI-assisted output should be reviewed by someone who takes responsibility for it.
• Protect personal information. Never enter identifiable client, donor or volunteer data into tools that are not approved for it.
• Be transparent where it matters. Donors and communities deserve honesty about how their information and stories are handled.
• Start with a problem, not a product. Adopt AI where it relieves a genuine burden, and measure whether it does.

A human technology after all

The deepest work of the nonprofit sector — building trust, sitting with people in difficulty, deciding what is fair — remains irreducibly human. The best use of these tools is to protect time for exactly that work. If artificial intelligence gives our sector more hours for human presence, it will have served us well.