Executive Summary

The integration of Artificial Intelligence (AI) and Machine Learning (ML) is considered to be a strategic imperative for humanitarian and developmental organisations, moving them from reactive aid to proactive data-driven interventions. Such technologies offer a paradigm-shifting opportunity to optimise the performance of institutions like the Egyptian Food Bank (EFB) by utilising AI for better decision-making.

This policy brief builds on desk-based research and a multi-stakeholder Webinar discussion on ‘’Utilising AI and ML to Assist in Optimising the Performance of Humanitarian Aid Institutions and Entities’’ (18 December 2025). The brief synthesises sector-wide evidence extracted from the discussions recommendations provided by significant stakeholders, lessons-learnt, and policy-relevant recommendations to guide responsible AI and ML adoption within humanitarian and developmental organisations.

Hereafter, the brief highlights key challenges related to data governance and responsible usage of AI (AI for Humanity), infrastructure and context-related challenges in Egypt, and ethical risks. This is emphasised while outlining practical lessons and actionable recommendations to be adopted by humanitarian organisations and development actors in Egypt.

I. Introduction: Strategic Context and Rationale

Data is quintessential for humanitarian organisations as they rely on it to guide decisions related to beneficiary targeting, logistics, supply chain management, and early warning and forecasting systems. Given the high multi-source volumes of data generated in humanitarian operations, traditional analytical ‘’static’’ methods are often insufficient and cost-prohibitive. Consequently, for organisations like the EFB, adopting AI and ML is not a mere technological upgrade, rather, a strategic necessity to achieve its mission more effectively and sustainably. Since traditional analytics methods will not suffice, AI will provide the tools to boost beneficiary targeting, forecasts food crises, and optimise the logistics and the overall operations of the EFB.

This strategic pivot aligns the EFB with the key national priorities in Egypt and the global ones as well:

·         National Alignment: Adopting ethical AI aligns with Egypt Vision 2030 and the Ministry of Social Solidarity’s (MoSS) move towards a digitised ‘’Unified National Registry’’.

·         A Paradigm Shift and a Leading Role: AI will offer the analytical power to support the ‘’Zero Hunger’’ mandate (SDG 2) by giving organisations like the EFB from simply ‘’feeding’’ to ‘’nutritional security’’. The latter being conducted by offering the ability to analyse health data, nutritional values and intakes of food i.e., iron, zinc, vitamins, minerals, and protein versus caloric needs nationwide.

II. Key Challenges in Applying AI in the Humanitarian Context

II.I Data Availability (Granularity and Cleanliness) and Governance:

·         Humanitarian organisations’ data face the challenge of fragmentation and, in order to integrate AI and ML algorithms, are required to be clean and machine-readable in order to avoid generating inaccurate predictions with regards to beneficiaries needs.

·         Interoperability between organisations is a challenge as data standards may be inconsistent across organisations.

·         With regards to privacy, consent, and protection of data of vulnerable populations, the sensitive data often raise concern on the governance question of AI and ML integration.

·         Algorithmic bias: AI models operating on biased historical data could amplify existing systemic inequalities which may affect the resource allocation.

II. II The Ethical Risk of the ‘’Black Box’’:

·         For the EFB and similar humanitarian institutions, transparency is paramount. Usually, deep learning models do function as ‘’black boxes’’ meaning that the logic behind their decisions is somehow opaque. Consequently, if an AI or ML model does remove a family from the eligibility list due to a correlation of data points, the humanitarian workers/staff should answer the ‘’why’’ to beneficiaries, partners, donors, and community leaders to maintain trust. Hence, AI data analysis and the decisions provided by it should not remain unchecked.

II. III Technical and Operational Challenges:

·         Limited access to funding to establish a robust digital infrastructure and computing power.

·         Static data might fail to capture ‘’transient poverty’’ and families becoming at high risk of vulnerability, in this case food insecurity and hunger, by sudden shocks e.g., unemployment, inflation, and health crises.

·         A shortage of local expertise of data science, particularly with regards to ML and AI usage, AI ethics, and data governance may hinder the successful implementation and sustainability of initiating a long-term project for utilising AI in humanitarian aid.

 

 

III. Lessons Learnt (Webinar Insights)

After highlighting on the challenges during the discussion with the speakers, they provided their insights on how humanitarian organisations can best integrate and utilise AI and ML in their operations tandemly while maintain AI ethics.

The session emphasised on the following:

1.      AI as decision-support tool, not an ultimate decision-maker: Once used to augment human expertise, AI delivers the greatest value. Rather than replacing human expertise, human decision-makers still become the ones responsible for ‘’contextual judgment’’ and accountability.

2.      Contextualisation as a priority: AI solutions, recommendations, and outputs must be locally relevant. Simply put, models must be trained on regional data validated by the local experts through their understanding of the unique socio-economic contexts of Egypt.

3.      Adopting an interdisciplinary multi-stakeholder approach: Collaboration is key for successful AI integration which means NGOs (i.e., EFB), research institutions (A2K4D and Mena Observatory), think tanks, governmental institutions, and the private sector should collaborate to ensure the balance of technical expertise, operational knowledge, and funding.

4.      Assessing Vulnerability and Poverty through proxy indicators: AI can be harnessed to analyse indicators such as: inflation, climate’s effect on crop and local production, unemployment. These indicators can be used to identify vulnerable groups in real-time and predict incoming crises, hence, the EFB’s intervention and other humanitarian institutions becomes pre-emptive.

5.      Embedding Responsible AI Principles: Ethical considerations and abiding by them are as essential as the effectiveness and fruitful outcomes that might be the result of using AI. Hence, they must be integrated from the design phase and are pillared on data governance frameworks for algorithmic transparency and data protection of vulnerable target groups.