NEMO of WFP
NEMO: How WFP Uses AI and NLP to Improve Humanitarian Data Management
How the World Food Programme explored AI and Natural Language Processing to improve the handling of large volumes of unstructured information — and what the experience teaches organizations about sensitive data, teams and human-centered AI.
Case Study at a Glance
Organization: United Nations World Food Programme (WFP)
Innovation: NEMO AI
Technology: Artificial Intelligence and Natural Language Processing (NLP)
Challenge: Managing large volumes of unstructured information
Reported result: Up to 15% reduction in call time
Reported capacity: Up to 50 times more cases handled within the same timeframe
1. The Challenge
WFP works with programmes that involve large numbers of people and generate substantial amounts of textual information. The WFP Innovation Accelerator notes that keeping track of this unstructured information and producing actionable insights can be challenging.
WFP identified Artificial Intelligence and Natural Language Processing as technologies with potential to improve efficiency in data collection and data-quality processes.
How can large volumes of unstructured information be processed more efficiently while keeping the information useful to the people who work with it?
2. The NEMO AI Innovation
NEMO is one of WFP’s frontier-technology innovations. The WFP Innovation Accelerator presents it as an example of how AI and NLP can be applied to improve the efficiency of managing unstructured information.
The published lessons learned focus not only on the potential benefits of the technology, but also on the practical challenges of implementing AI when sensitive information is involved.
3. Reported Results
Reduction in call time
WFP reports a reduction in the time required for calls of up to 15%.
Case-handling capacity
WFP reports that up to 50 times more cases could be handled within the same timeframe.
Evidence note: These figures are reported by the WFP Innovation Accelerator. The source does not provide a detailed evaluation methodology for independently verifying these figures.
4. Sensitive Data: A Major Implementation Challenge
One of the most important lessons from NEMO concerns the additional requirements created by sensitive data.
WFP states that privacy and security requirements can create significant additional work when developing AI solutions. According to the Innovation Accelerator, handling sensitive data can extend the Proof of Concept (PoC) or Minimum Viable Product (MVP) phase by approximately one to two months.
Sensitive data should be treated as an important implementation consideration rather than something addressed only after an AI solution has been developed.
5. The Importance of the Right Team
WFP also highlights the importance of a multidisciplinary team. The Innovation Accelerator identifies four types of expertise that can contribute to an effective AI innovation team:
- AI engineers
- AI practitioners
- Business analysts
- Project managers
WFP’s stated rationale is that this combination helps ensure that innovations meet technical requirements while also aligning with broader organizational goals and project requirements.
6. Human-Centered Design
WFP’s experience also emphasizes that technical performance is only one part of successful AI implementation.
The Innovation Accelerator specifically highlights the importance of designing the human-machine interface and processes carefully, including how recommendations and probabilities are presented to users.
AI outputs need to be presented in ways that users can understand and use. User-centered design is therefore part of the implementation process, not simply a visual-design consideration.
7. How M&E Professionals Can Adapt This Approach
WFP’s NEMO experience provides a useful model that M&E and MEAL professionals can adapt to their own information-management challenges. This does not mean reproducing NEMO itself. Rather, the transferable idea is to identify a high-volume information process where AI and NLP could reduce manual effort while keeping people involved in reviewing and using the outputs.
1. Start with a real M&E problem
Identify a process involving large amounts of unstructured information rather than starting with a particular AI tool.
2. Identify information that can be processed with NLP
Examples could include qualitative feedback, interview transcripts, open-ended survey responses or narrative programme information. These are adaptation examples, not applications specifically reported for NEMO.
3. Define the human role
Decide what AI should assist with and what remains subject to professional review and judgment.
4. Address sensitive data before implementation
Determine whether the information contains personal, confidential or otherwise sensitive data and establish the required safeguards before building the solution.
5. Measure the operational change
Establish baseline measures such as processing time, workload or case-handling capacity so that the effect of an AI intervention can be assessed.
A practical M&E adaptation model:
Identify the information problem → assess the data → define the AI-assisted task → establish human review → address privacy and security → test with users → measure operational results.
This adaptation follows the implementation lessons explicitly identified by WFP: focus on the information challenge, account for sensitive data, use multidisciplinary expertise, and design the interaction between people and AI carefully.
8. What M&E Professionals Should Not Assume
The WFP source does not provide enough information to conclude which specific AI model NEMO uses, how accurate the model is, or how the reported operational figures were calculated.
Therefore, this case should be understood primarily as a documented example of AI implementation and organizational learning, rather than as a complete technical evaluation of an AI model.
9. Key Lessons
- Start with the problem, not the technology.
- Plan for sensitive-data requirements early.
- Combine technical and programme expertise.
- Design AI around the people who will use it.
- Measure whether AI actually improves the underlying process.
Key Takeaway
NEMO shows that implementing AI in humanitarian and M&E-related environments is not simply a matter of selecting an AI model. WFP’s experience points to a broader implementation equation: data + privacy + people + workflow + user-centered design + measurable operational value.
Primary Source
WFP Innovation Accelerator
WFP innovation lessons learned 2023: Navigating new frontiers
Read the WFP Innovation Source
This case study is based primarily on the WFP Innovation Accelerator’s published lessons learned. Adaptation suggestions for M&E professionals are clearly identified as recommendations rather than claims about how WFP implemented NEMO.
