Recorded at the AI for Good Global Summit in Geneva, 2026.
A theme that came up all day: AI doesn’t do much without people. The clearest example was a humanitarian mapping project in Ghana, where local knowledge fixed what a global model got wrong.
Key takeaways
- Partners like Google and McKinsey bring the technology and the logos. Humanitarian workers do the grounding work: building trust with communities so aid is deployed well.
- The CSR partnership and the logo get the press release. The trust-building on the ground is what makes the project work.
- A nonprofit in Ghana found a building-detection model couldn’t recognize buildings in Accra, because it had been trained on Global North rooftops, materials and density.
- People on the ground taught it what a house looks like there, working with Google.org to improve the dataset, and it keeps getting better.
Try this with your team
Where is your organization the “people on the ground” in a partnership with a larger tech or funding partner? Is that knowledge credited, and is it shaping the tools?
Going further
If you’re weighing whether a tool fits your context, try the free Can I Use AI for This Workflow? check.
More from this series
Transcript
Lightly edited from YouTube’s automatic captions for readability. Wording and order are the speakers’ own.
Just wanted to update on what I did today at AI for Good. One of the big themes that kept coming up was that the fancy AI doesn’t really do anything without humans, and those could be humans you see in the room or not.
The partnerships there, like Google and McKinsey, are really not worth anything without the humanitarians on site. They’re doing the unglamorous work, but they’re doing the grounding work for these organizations. They’re building trust with communities. The CSR and the logo is what gets the press release.
It was about the deployment of humanitarian aid in Ghana. There was a nonprofit that noticed they couldn’t get a good read on what was a building in Ghana, because the buildings Google Maps had been training on were very much Global North structures. The rooftops and the materials, the density of where buildings were, just didn’t match what it looks like from above in Accra.
So people on the ground had to teach what a house looks like in the Global South, and now it just keeps getting better and better. They cooperated with Google.org to improve the dataset.