Our Projects

Microsoft Generative AI

In November of 2023, GIEVA became an awardee of the Generative AI Skills Challenge!

Awards were given to five organizations around the world working with generative AI to accelerate digital inclusion and skills advancement and acquisition for workers from historically marginalized populations. Global integrated Education Volunteers Association (GIEVA) was selected from a competitive pool of nearly 600 applicants in 93 countries. The award provides us with funding, technical resources, technical guidance from Microsoft AI experts, and data training.

The generative AI project was initiated to develop AI skills of the underrepresented women in Nigeria. We tackle the identified challenges through AI skilling, training and mentorship to empower 600-1000 low-medium income business women in Northern Nigeria, precisely Kaduna and Plateau States. This is to leverage artificial intelligence to foster economic growth and inclusivity among women entrepreneurs often marginalized in digital advancements, and thus, bringing them into the digital ecosystem. Also, it   fosters a demographic that is well equipped to contribute to the development and responsible use of the application of AI technologies. By creating a unique-tailored AI training and skilling curriculum, an AI solution that uses LLM’s to create digital assets for sharing over social media platforms was created. This increased their sales, and ability to develop compelling online product profiles and sales assets which optimized the sharing over WhatsApp and social selling platforms. 

The AI solution built, considered the women’s limited digital literacy or fluency in English – the dominant online language. In addition, the AI solution created assets via low-end smartphones thus avoiding the need for a laptop or desktop. This was achieved through careful user tests with marketplace sellers in Nigeria to highlight the specific challenges they face while using low-end android devices. These challenges include budget for data, the need for  frequent battery charging, and their digital and linguistic usage patterns- limited English language that might confound LLMs. 

Our Project methodology was four-fold:

Firstly, to conduct a comprehensive research on low-medium income business women in Northern Nigeria, to determine their knowledge or lack of Gen AI tools, bias, if any,  and limitations to its use for economic advancement. Secondly, to use the data from the research to develop an AI solution,  and a training curriculum targeted at low-medium income business women. Thirdly, to train and upskill 600-1000 business women in the use of Gen AI tools to scale up their business and bring them into the digital economy. And fourthly, design training personas of the marginalized women based on the research reports.

Goals and objectives, with results achieved

Project goals are:

Goal One: Assess and perform a comprehensive study to understand the impact of AI tools and processes on boosting sales and profits for businesses led by women in low-to-medium income communities in Nigeria.

Results: A total of 1632 women-owned businesses in Kaduna and Plateau States were interviewed and data collected based on the questionnaires administered at markets in the following locations:

Plateau: Kabong, Jenta Adamu, Farin Gada, Building material roadside, Rukuba Road, Basa Market, Angwan Rukuba, Tudun wada Market, Jenta Adamu Market.

Kaduna: Sabo Market, Barnawa, Narayi, Kawo, Tirkaniya, and Angwan Muazu.

The research report findings include;

• About 25% of the women risk digital exclusion and they do not have any digital devices.
• 40% of the women lack technical skills as reasons for not engaging in digital activity.
• About 7 out of 10 women have never heard about AI or Generative AI.
• More than 80% of the women are willing to be trained in digital skills and Generative AI. 

Goal Two: Create an AI solution tailored for businesses operated by women in low-to-medium income areas, designed for use in both online and offline marketplaces as well as on social media platforms.

Results: The Development of a functional LLM AI solution Jumm.ai based on results of findings from preliminary research report. The application considers low literacy levels, addresses the language barrier which might confound the LLM by incorporating Hausa language data sets via the language translation feature of the application.  Other features include image generation, conversational chatbot, and WhatsApp integration. 

The Jum.ai application was designed to use minimum data and low battery consumption. This forms a central part of training and skilling the women.

Goal Three: Establish an AI training and mentorship program to build AI learning capabilities among the targeted group of women and tailor LLM resources and models to align with the unique cultural, linguistic, ethical, and economic contexts of low-to-medium-income communities.

Results: Establishment of training hubs in Kaduna and Plateau states where 1043 women were trained in the use of generative AI tools such as chat GPT, Canva, Jumm.ai, and meta ai to generate images, create fliers, get tailored advice on how to improve their seller ranking, and financial literacy and tools .

The development of a tailored and adaptive curriculum and manual targeted at the identified needs of the business women, was used for training instruction.

Objectives: Bring women entrepreneurs from low-medium income demography into the digital economy, train and up-skill women entrepreneurs on the use of digital capabilities to create sustainable digital livelihoods.

Status of Milestones and Outcomes:

1. Preliminary Research Report findings: facilitated by Department of computer science Uni. of Jos

Status: Achieved

The results of the landscape analysis, and reference materials for subsequent research in related fields are available on GIEVA website. 

2. Prototype development: Jumm.ai 

Facilitating Organization: Zeustek ICT solution

Status: Achieved: The AI tool (jumm.ai) is tailored to low-medium income business women, designed for both online and offline marketplaces as well as social media platforms. Jumm.ai.

Key features of the Jumm.ai app are;

• Multilingual capability: A unique feature of jumm.ai is its ability to converse in Hausa language, a widely spoken language in Northern Nigeria, this feature allows ease of use by all women regardless of their educational background.
• Conversational chatbot: Jumm.ai includes a chatbot that interacts with users in a humanlike manner in English and Hausa languages, providing instant responses to users and helping business owners with customer inquiries as well as generating instant content for their social media pages, thereby increasing their online presence.
• Image generation: Another key feature is the image generation. Jumm.ai generates custom images tailored specifically for users based on their description and prompts making each image unique to the user.
• Voice translator: This feature allows jumm.ai to translate spoken languages in real time thereby breaking language barriers across different languages.
• AI whatsapp integration: Another unique feature of the Jumm.ai is its ability to provide instant replies to customer inquiries on whatsapp acting as a virtual assistant to business owners, thereby improving customer service efficiency.

3. Testing the prototype: 

Implementing organization: Colab Innovation Hub and Zeustek ICT solution

Status: Achieved: Our team identified and recruited 67 business women from the local community to participate in a pilot test of our newly developed AI prototype jumm.ai at various stages in its development. These women were chosen based on their entrepreneurial spirit and their businesses’ potential for growth through AI integration. 

The outcomes of user tests, exceptional scenarios, data usage metrics, findings on user limitations, and hardware constraint results as resource material for future reference.

4. Development of curriculum and training manual 

Implementing organization: Colab Innovation Hub

Status Achieved: In the creation of our training curriculum for low-income women, we employed a persona-driven approach to enhance its relevance and efficacy. Detailed personas were meticulously developed, representing the diverse characteristics of our target audience. These personas encompassed various aspects such as socioeconomic backgrounds, educational levels, cultural contexts, and personal ambitions. This strategy enabled us to craft a curriculum that was not only highly customized but also resonated deeply with the participants’ real-life experience.

The creation of the training curriculum for low-income women, the target audience, involved a tailored approach. Recognizing their unique challenges and needs, the curriculum was crafted with a focus on practical, applicable skills that could directly enhance their livelihoods. Extensive research and consultations informed the content, ensuring it was accessible, relevant, and empowering. Special attention was given to overcoming barriers such as limited literacy and technological access. The curriculum included interactive and hands-on learning experiences, supported by resources suited to their contexts Manual.

5. Training and skilling 

implementing organizations: All consortium members

Status: Achieved: Business women in the low-to-medium income range in Kaduna and Plateau learnt how to skillfully craft localized prompts that are precise, contextually relevant, and allow the generative AI tool to produce optimal output. Testimonials videos, training photos and training administered in Hausa language. GIEVA.org

6. Collaboration and Partnerships: 

Status: In progress.  Work with data.org to convey impact to a global audience via story of impact published on the data.org site.

Partnership with microfinance institutions, particularly Safe Haven Microfinance Bank for implementation of intervention funds to identified beneficiaries and other long-term financial stability for the businesses involved such as offering low-interest loans. On-going discussions and engagement with government agencies specializing in women business or technological development, such as Maryam Babangida National  Center for Women Development, Ministry for Women Affairs- Katsina State and National Information Technology Development Agency (NITDA).

These local collaborations are strategic moves designed to ensure the project’s sustainability and growth plan.

7. Technical infrastructure

Status: Achieved: Data from the research team was collated and analyzed in the development of the prototype (jumm.ai) that utilizes language model technology to further prepare the app development phase. 

We identified edge cases that could benefit the project and addressed issues related to data consumption and hardware malfunctions through further research and data collection. A team of software engineers gathered the necessary AI tools and moved to the development stage. This resulted in the development of the first of its kind prototype which was hosted on Microsoft Azure ML platform, immediately making it available to the team members for extended testing which was an iterative process.

8. Communication: Revisit the communities to identify participants who will benefit from the project’s interventions. We shall publish the outputs on the public facing websites of GIEVA, Colab, Zeustek ICT and Solutions

Status: in progress. Post training survey’s ongoing, Training ended June 28, 2024 in some consortiums. assessing the impact of the skills acquired on their businesses may take a few weeks, the plan is to provide ongoing support, follow up and evaluate 12 weeks post training. 

9. Project Media and publicity

Status: Achieved:

1. Awareness campaign: carried out in Kabong market and Gada biu Market, Jos, Plateau State on April 20,2024
2. Hosting talks shows on radio stations to raise awareness and announce commencement of Gen AI training 
3. End conference: GIEVA hosted a one-day conference in Abuja, Nigeria May 22, 2024. The theme of the conference was ‘Accelerating Digital inclusion and skills Advancement & Acquisition in Nigeria’.

In attendance were stakeholders from Government, Civil society, technology, Persons with Disability leaders, and business women. 

 Details of the conference were published in Nigerian Newspapers, reputable TV stations: 

10. Publishing in a journal/ Open source data: Status: Pending: Preparing the manuscript for publication

is in concluding stages. The publication is on findings of our research, the generative AI solution built to

support the women and the resultant impact of training these women on their businesses. Open source

documentation for the Jumm.ai is ongoing, the application is currently hosted on our private app. once

documentation is completed, it will be available on JUMM.AI APP.

Demographic Information and Metrics 

A total of 1,043 women were trained in Plateau and Kaduna state by GIEVA, Zeustek, DCS and Colab. 100% our project demographic are women from historically marginalized communities in Northern Nigeria. According to the global data index in 2018, an average of 62.5% of households in Kaduna are poor, with an international wealth index (IWI) value under 70. This suggests a large percentage of people as low-income earners. Almost the same can be said of the Plateau state. Communities in Kaduna and Plateau states  are affected by ongoing conflicts, such as those instigated by Boko Haram, banditry, kidnapping, and religious crisis leading to displacement and insecurity, high levels of poverty with limited or non existent basic amenities such as education, healthcare, electricity and portable water, and densely populated areas are prevalent factors that classify these communities as marginalized. Another key factor is that Educational attainment is generally lower in Northern Nigeria compared to other regions, particularly among women and girls because of  adherence to traditional and religious practices, which can sometimes hinder progress in areas like women’s rights and education.

Access to and knowledge about basic technology is still a thing, around 27% of the women do not have smartphones, highlighting a disparity in access to mobile technology which might be due to various factors such as affordability, digital literacy, or personal preference. 74% of the business women surveyed do not have knowledge about Gen AI.

Of 107 surveyed training participants 71.2% have social media accounts, 28.8% do not use social media at all, Women without social media accounts were successfully onboarded and taught how to use social media. 25.3% of the participants have used social media to advertise their businesses while  64.71% have not used social media for business advertising, These insights helped us to understand tailored intervention we  provided to them through training and skilling. A total of 17 facilitators participated in the Gen AI training deliberate effort was made at ensuring recruitment of women AI trainers.

Certificates of participation were awarded to trainees upon successful completion of digital training.

Impact Stories and Use Cases:

AI Literacy

A pre-training survey result from 107 respondents at Kabong Market showed that a staggering 86.8% of the women have never heard of or used Generative AI tools, while a pre- training survey at Zeustek training hub, showed that over 60% of trainees had no knowledge of what AI is or how impactful it will be to their businesses, by the end of the training over 1000 business women learnt to prompt properly to get customized results, and create fliers and business logos using AI tools, they also learnt to generate content leveraging on AI’s intelligence.

Business Growth: Preliminary post training surveys indicate that 85% of the users reported an improvement in business visibility and a 60% increase in sales after using the platform for a few weeks. 

Social media onboarding and engagement:

A large number of the women have never used any social media channel to market their goods, so we designed a training module on “Building an Online Presence” which featured a practical session on how the women can use WhatsApp Business and Facebook Marketplace. The women share how they have been able to start exploring digital marketplaces and selling platforms. We also created WhatsApp groups for each training cohort , where the women participate by sharing content such as logos and images generated with our Jumm.ai prototype and Meta AI.

Ease in Content Creation: 89% of participants reported ease in content creation thanks to jumm.ai.

Gen AI Awareness and interest

GIEVA embarked on a Gen AI awareness campaign at Kabong and Gada Biyu markets  in Jos, April 20, 2024. The campaign featured door to door market stall visits to raise awareness on Gen AI, its benefits to their businesses and the introduction of our AI solution (jumm.ai) to the market women. These campaign sessions spiraled into a major township awareness project. Ten days after the campaign, one of the women leaders at Kabong market (Mrs. Florence) requested the GIEVA team to train the entire market women on the importance of Generative AI, since the Jumm.ai application had sparked great interest among the local women. The campaign had actually taught the women (and especially Mrs. Florence) how to download the Jumm.ai prototype and create adverts, and were able to post pictures online and get customers visiting their shops. As a result of their enthusiasm, the market management offered the market secretariat as a training venue for the project. 170 women were trained at the Kabong market in Jos- Plateau state.

Community Feedback: The feedback from the community has been overwhelmingly positive, with many users highlighting the ease of use and the practicality of the tools provided, community leaders also canvas support for Gen AI project  in local language.

Government interest:

At the GIEVA Gen AI training opening ceremony in Jos, the Government of Plateau state through the commissioner of science and technology Prof Goselle Obed Nanjul, announced plans to support training of 5,000 women on Gen AI in Plateau state. 

The commissioner for women Affairs in Katsina state (Hajiya Zainab Musawa) also requested a proposal to train women in Katsina state on the use of Gen AI tools for business empowerment. GIEVA has submitted a proposal which is under review.

Individual  testimonials

Mrs. Rifkatu Jerry, a participant at Kabong market training, did  not have a smartphone, but as a result of the training program has seen the need to purchase a smartphone, because of what she has just learned on the benefits of using digital tools.

Ammi Baraya, Business Name: Barammi Couture,  Instagram: @Barammi, Facebook:  Barammi Couture  said ‘the gen AI training program exceeded my expectations, providing a thorough understanding of digital marketing, business fundamentals, social media marketing, dropshipping, graphic design, and the applications of Generative AI in business. 

The training had a profound impact on my business, enabling me to; Automate manual tasks, freeing up time and increasing productivity, Streamline creative idea generation using the Jummai tool and unlock new design    possibilities. I am also able to personalize customer interactions, fostering loyalty and retention through effective online engagement and Author the ebook “African Style Kit,” showcasing my expertise and passion for fashion’.

Challenges and Lessons Learned: 

Challenge 1: Prototype development delay: Development of the AI prototype (Jumm.ai) took longer than anticipated, this delayed commencement of training.

Approach Taken: inorder to meet our target to train at least 600 women, all consortiums engaged in recruitment and training of women.

Challenge 2: Refusal to divulge information: During the survey, we noticed that low-income women are unwilling to give out information about themselves and their businesses, especially if the data collector is male. They expected some form of payment for their information. 

Approach taken: To address this, we paired male Research Assistants with female ones and explained to the women that the exercise was meant to help their businesses. This solved the problem.

Challenge 3: Google Play Store approval: At this time, updates on the jumm.ai have to be shared via  apk directly to the trainees to download. This is as a result of the pending approval of the  jumm.ai to be downloadable on  google play store.

Approach taken: sent and responded to  emails from google play store team on any documentation requested.

Challenge 4: Reluctance to participate in digital activities: Low-income women who are unwilling to participate in digital activities, they believe that technology is not for them but for those with knowledge and education. They generally feared embracing technology. 

Lesson: we learned that training these women changed everything and gave them confidence to use technology effectively.

Challenge 5: Language barrier: A large percentage of the women weren’t fluent in English. 

Approach Taken: We employed trainers who taught Generative AI in the Hausa language.

Lesson Learned: To build a community of women learning data science in diverse languages.

Challenge 6: Slow learners We had a number of slow learners, especially older women who were not very familiar with technology 

Approach:. Wait behind after classes and give them special attention, scheduling special days for them, and grouping the trainees in cohorts according to literacy and educational levels as obtained from pre-training survey data.

Lesson Learned: With patience and a more targeted training approach, the women were able to catch up quickly. 

Challenge 7: Difficulty in operating Smartphone, Some of the trainees could not operate their smartphones, or the software was outdated.

Approach: updating phones software and taking time to teach them basic use of functions on their phones.

Lessons learned: some women are very enthusiastic to learn, they borrowed phones they cannot operate to attend the training.

Challenge 8: Crowd control: initially, most women were skeptical about joining the training, we reached out to their kids, church leaders, and local government women leaders after the first few sessions of training, the number of women that turned out exceeded the expected projections thereby stretching the hub’s capacity. 

Approach: The trainees were divided into three batches Each batch had two streams, one from Monday to Wednesday and the other from Wednesday to Friday, with three daily sessions: morning, afternoon, and evening

Lesson: We learned to be deliberate when engaging low-income businesswomen. Community champions play a big role in women’s engagement.

Challenge 9: Distraction Many women with infants and toddlers brought them to the training sessions, often distracting the class. 

Approach: training hubs had to make provision for Creating a makeshift daycare for the younger children to sit, play and watch cartoons under the supervision of the staff while the mothers attended their sessions.

Challenge 10: Technical support: The credits on Microsoft Azure got exhausted at some point, which made the Jumm.ai application unavailable for use by the trainees. There was also the challenge of Low Hausa language resources available.

Approach:: The GIEVA consortium is committed to the iterative improvement of the jumm.ai application in areas of improving image generation, population of the prototype with culturally relevant images and localized data sets. As such, purchased additional credits to keep the application running.

Lesson: The Hausa language datasets need to be populated and fine tuned.