Shahzad Asghar: Artificial Intelligence in UN Operations
How I have applied artificial intelligence and data analysis to humanitarian operations in the UN system, and what those projects produced.
Published 2024-07-02 · Updated 2026-10-23 · By Shahzad Asghar
This page summarises how I have applied artificial intelligence and data analysis to humanitarian operations in the United Nations system, and what those projects produced.
Current role
I work as Senior IT Strategy and Portfolio Officer at UNRWA, where my work covers IT strategy, portfolio management and the governance that decides which technology investments an operation actually makes.
Data and digital work at UNHCR
From 2018 to 2024 I led the Data Analysis Group at UNHCR Jordan, a team of 12, as part of a wider multi-year digital transformation programme. The unit existed to put reliable figures in front of the people making operational decisions, which in a refugee response means registration, appointments, assistance targeting and service delivery rather than reporting for its own sake.
The work included:
UNHCR's first global interactive voice response appointment system, which let refugees book an appointment by telephone, in their own language, without travelling to an office or needing an internet connection. It was used by more than 700,000 refugees, it reduced registration service time by 75 percent, and it was replicated across five country operations.
A forensic data integrity analysis of the refugee registration database, which flagged 165,000 records as possible duplicates for controlled investigation and identified 63,000 missing contact records. The flags went to human investigation rather than to an automated merge, and 27,000 individuals were subsequently offboarded from assistance programmes.
Operational dashboards that tracked projects against their targets, so a problem surfaced during a programme rather than after it.
Moving analysis workloads onto cloud infrastructure, mainly Azure services, which removed a standing constraint on how much data the unit could process.
The data integrity work involved no machine learning of any kind, and it remains the most consequential thing on that list. Every system built on that database would have inherited its errors.
Earlier roles
My career began in 2005 at Aga Khan University Hospital, then moved to the Canadian International Development Agency in Pakistan, where I worked on public health programmes addressing HIV and AIDS and polio eradication. That work was built on survey and immunisation data at national scale, and it is where I learned that the analysis is the easy part. Getting a field team to trust and act on a number is the hard part.
My first United Nations role was with UNICEF in Kenya in 2014. I have also worked with UNOCHA, and after UNHCR I headed data and digital solutions at UNESCWA before joining UNRWA.
AI governance and data protection
Most of what I write about now concerns governance rather than models. Humanitarian data describes people who cannot easily refuse to be recorded, and who carry real risk if the data is wrong or leaks, so the questions that matter are about consent, retention, access and what happens when a system is confidently wrong.
The positions I hold and argue for:
Data governance precedes AI governance. You cannot govern a model built on a database nobody has audited.
Data protection is a design constraint, not a compliance step applied at the end.
An AI system used in an operational decision needs a human review path, and that path has to be usable under field conditions.
Tools built for well-connected offices fail in the places that need them most, so bandwidth and device assumptions belong in the design from the start.
Teaching
I have run courses on data visualisation and cloud computing for UN teams. The emphasis is on turning a dataset into something a decision maker can act on, because a technically correct chart that nobody reads has changed nothing.
Speaking and participation
My work has been presented at international forums, and I have been invited to take part in high-level discussions on the future of AI within the United Nations system. What I contribute to those rooms is the operational view: what a system does once it meets a field office with intermittent connectivity and a caseload that does not pause.
Writing
I publish what I learn from these projects, including the parts that did not work. The articles on this site cover agentic AI in humanitarian operations, data protection, and building systems where the infrastructure cannot be assumed.
Written by Shahzad Asghar — Senior IT Strategy and Portfolio Officer at UNRWA, with 20+ years building AI and data systems across UNHCR, UNICEF, and UNOCHA. His team built UNHCR’s first global IVR appointment system, serving 700,000+ refugees. He created the Last-Mile AI Framework. Read more about this UN AI expert
Related articles
- Open Source and AI in Digital Transformation: A Practical Playbook for Organizations (2026-10-04)
- Three Lessons from the OIOS Digital Transformation Evaluation for the AI Era (2026-09-18)
- Why I Am Using McKinsey's Vizro (And Why You Should Too) (2025-12-03)
- Artificial Intelligence’s Transformative Impact on Industry and Economy (2025-04-01)