Microsoft
has announced Project Gecko, a new artificial intelligence initiative designed
to address performance gaps in generative AI for communities underrepresented
online, particularly speakers of low-resource languages. The project was
unveiled on Tuesday, 18 November 2025, and marks a major step in the company’s
efforts to create more inclusive and culturally grounded AI systems.
According
to Microsoft, many existing AI models perform poorly in regions where
linguistic diversity is high and data availability is low. These shortcomings
have contributed to lower adoption of AI tools in countries where local
languages dominate, even when internet access and GDP levels are taken into
account.
Project
Gecko aims to bridge this gap by building cost-effective, adaptable AI systems
tailored to communities across Africa, Asia, and other parts of the global
majority. The effort is being led by Microsoft Research Africa in Nairobi,
Microsoft Research India, the Microsoft Research Accelerator in the United
States, and partners across agri-tech, philanthropy, and academia. The
initiative focuses on developing AI that can communicate in local languages,
incorporate culturally relevant knowledge, and support multiple interaction
modes including text, speech, and video.
At
the center of the project is the MultiModal Critical Thinking Agent
(MMCTAgent), a new AI system capable of analyzing speech, images, and video to
deliver context-rich, locally grounded answers. MMCTAgent is now available
through Azure AI Foundry Labs, with open-source resources published on GitHub.
Ashley
Llorens, Corporate Vice President and Managing Director of the Microsoft
Research Accelerator, said the project reflects Microsoft’s broader mission to
build AI systems that empower people in all parts of the world. “Building AI
systems from the ground up shaped by the knowledge, languages, and modalities
of the global majority yields more innovative, useful solutions for a great
number of people,” he said.
Agriculture as the Starting Point
While
Project Gecko is expected to expand into healthcare, education, and retail,
Microsoft chose agriculture as the initial focus due to its economic
significance in countries like Kenya and India. Agriculture employs millions of
smallholder farmers in these regions, many of whom operate on less than five
acres of land.
However,
linguistic diversity and cultural nuances often limit the effectiveness of
existing AI tools in this sector. Farmers regularly navigate multiple
languages, rely on oral communication, and face bandwidth and device
constraints. Studies conducted by Microsoft found that many agricultural apps
provide incomplete or inaccurate information because their underlying models
were primarily trained in English.
To
address these challenges, Project Gecko builds on Digital Green’s FarmerChat—a
speech-first assistant used widely by extension workers. Digital Green has
amassed more than 10,000 agricultural videos in over 40 languages, but until
now, these resources were difficult to access due to language barriers.
With
Project Gecko’s enhancements, farmers can ask questions verbally in their local
languages and receive responses in text, audio, and video. The system can even
direct farmers to exact timestamps in community-generated videos where
solutions are demonstrated.
Tanuja
Ganu, Director of Research Engineering at Microsoft India, emphasized the
importance of linguistic precision in agriculture. “Agriculture has very
specific terms, which may change from language to language, and even district
to district. All those domain-specific nuances need to be understood,” she
said.
Locally Trained Speech Models
Microsoft
researchers found that many farmers prefer voice interactions, prompting the
team to develop new automatic speech recognition, text-to-speech, and
translation models for languages with previously limited datasets. These tools
are being built using small language models (SLMs), optimized to run
efficiently on low-cost devices common in rural communities.
The
team has collected more than 3,000 hours of crowd-sourced Kenyan speech,
expanding support to Swahili, Kikuyu, Kalenjin, Dholuo, Maa, and Somali. A
public leaderboard is also being developed to evaluate AI performance in
African languages.
Real-world
insights from more than 130 farmers have informed new features including
clarifying prompts, actionable recommendations, and peer-to-peer
knowledge-sharing tools.
Next Steps
Microsoft
says Project Gecko is part of a broader effort to create equitable and
responsible AI systems that reflect global diversity. The company is preparing
to release a multilingual playbook offering guidance to developers building
localized AI tools across essential sectors.
“Our
goal is to ensure that the next generation of AI is not only powerful, but also
globally inclusive, culturally relevant, and shaped by the communities it aims
to serve,” said Ganu.
Project
Gecko is expected to expand further as the company continues its research and
partnerships across the global majority.
