Afrihealth Optonet Association (AHOA) - CSOs Network
Publicado el 05/11/2026
A Rejoinder to the UNEG Evaluation Practice Exchange Takeaways on AI in Evaluation
- By Dr. Uzodinma Akujekwe Adirieje, FAHOA
CEO and Programmes Director, Afrihealth Optonet Association (AHOA)
I thank the Evaluation Offices of IFAD, WFP and GEF for convening this exchange and for sharing the six takeaways so openly. I write from civil society in Africa, where evaluation is often done with limited resources and the cost of getting AI wrong falls on communities. I agree with the direction of the six messages. I want to press on four points.
a. "DOING MORE WITH LESS" MUST NOT BECOME "EVALUATING LESS."
Takeaway 1 is right that AI can shift effort from mechanics to meaning. In low-resource settings, though, efficiency gains tend to be absorbed as budget cuts. If AI-assisted desk reviews and report drafting become the reason to shrink field budgets, community consultations and evaluator fees, the promise is reversed. The capacity AI frees must be reinvested in interpretation, ethics and relationships, and commissioners should say so in their budgets.
b. HUMAN INSIGHT DEPENDS ON WHO THE HUMANS ARE.
Takeaway 3 makes human insight a defining feature of quality, and I agree. The human in the loop must also come from the context being evaluated. A verifier who cannot read the local language, or who has never sat with a community, cannot validate an AI output about that community. I have seen this as Chair of the Community Advisory Board for the VERDI Mpox study in Nigeria’s Bayelsa, Lagos and Rivers States. Legitimacy there came from people accountable to the participants. AI can support that work, but it cannot supply it.
c. LOCAL RELEVANCE MUST INCLUDE LOCAL OWNERSHIP.
Takeaway 5 calls for co-design, locally relevant data and sustained capacity. I would go further. National and civil society evaluators should co-design and co-own AI-supported methods and not only be trained in tools built elsewhere. Otherwise "capacity building" repeats the dependency it claims to end. Multilingual resources matter, because mainstream tools perform unevenly across African languages. Data sovereignty matters too, especially for refugees and displaced people, whose data is the most sensitive and who are the most easily left out of the evidence base. I raised these themes in my EvalforEarth blog, "Reimagining Evaluation for Climate Action in Africa: Indigenous Knowledge, Inclusion, Fragile Contexts and Artificial Intelligence."
d. GOVERNANCE OBLIGATIONS MUST RUN IN BOTH DIRECTIONS.
Takeaway 6 asks consultants and service providers to disclose, verify and use AI responsibly. I accept this fully, and AHOA holds itself to it. We disclose AI use, verify every output, protect data and keep a named human accountable for each conclusion. The same standard should apply to commissioners, who should also disclose how AI shapes their evidence mapping and commissioning decisions. Civil society networks should help design these governance rules and not only comply with them.
CONCLUSION
We hold that artificial intelligence must serve equity, justice and sustainable development and never replace human dignity. The six takeaways point the same way. What remains is to make them binding: in terms of reference, in budgets, in contracts and in who sits at the table. AHOA offers its network of more than 3,000 organisations across 113 countries, together with its experience in community engagement and M&E, to help turn these takeaways into practice.
Dr. Uzodinma Akujekwe Adirieje, FAHOA
CEO and Programmes Director, Afrihealth Optonet Association (AHOA)
2026 EvalforEarth Champion | President, African Refugees Council (ARC)
RE: Evaluation under pressure : Strengthening evidence use, accountability and learning amidst funding constraints (Joint EvalforEarth-ALNAP discussion)
Nigeria
Dr. Uzodinma Akujekwe Adirieje
CEO
Afrihealth Optonet Association (AHOA) - CSOs Network
Publicado el 05/11/2026
A Rejoinder to the UNEG Evaluation Practice Exchange Takeaways on AI in Evaluation
- By Dr. Uzodinma Akujekwe Adirieje, FAHOA
CEO and Programmes Director, Afrihealth Optonet Association (AHOA)
I thank the Evaluation Offices of IFAD, WFP and GEF for convening this exchange and for sharing the six takeaways so openly. I write from civil society in Africa, where evaluation is often done with limited resources and the cost of getting AI wrong falls on communities. I agree with the direction of the six messages. I want to press on four points.
a. "DOING MORE WITH LESS" MUST NOT BECOME "EVALUATING LESS."
Takeaway 1 is right that AI can shift effort from mechanics to meaning. In low-resource settings, though, efficiency gains tend to be absorbed as budget cuts. If AI-assisted desk reviews and report drafting become the reason to shrink field budgets, community consultations and evaluator fees, the promise is reversed. The capacity AI frees must be reinvested in interpretation, ethics and relationships, and commissioners should say so in their budgets.
b. HUMAN INSIGHT DEPENDS ON WHO THE HUMANS ARE.
Takeaway 3 makes human insight a defining feature of quality, and I agree. The human in the loop must also come from the context being evaluated. A verifier who cannot read the local language, or who has never sat with a community, cannot validate an AI output about that community. I have seen this as Chair of the Community Advisory Board for the VERDI Mpox study in Nigeria’s Bayelsa, Lagos and Rivers States. Legitimacy there came from people accountable to the participants. AI can support that work, but it cannot supply it.
c. LOCAL RELEVANCE MUST INCLUDE LOCAL OWNERSHIP.
Takeaway 5 calls for co-design, locally relevant data and sustained capacity. I would go further. National and civil society evaluators should co-design and co-own AI-supported methods and not only be trained in tools built elsewhere. Otherwise "capacity building" repeats the dependency it claims to end. Multilingual resources matter, because mainstream tools perform unevenly across African languages. Data sovereignty matters too, especially for refugees and displaced people, whose data is the most sensitive and who are the most easily left out of the evidence base. I raised these themes in my EvalforEarth blog, "Reimagining Evaluation for Climate Action in Africa: Indigenous Knowledge, Inclusion, Fragile Contexts and Artificial Intelligence."
d. GOVERNANCE OBLIGATIONS MUST RUN IN BOTH DIRECTIONS.
Takeaway 6 asks consultants and service providers to disclose, verify and use AI responsibly. I accept this fully, and AHOA holds itself to it. We disclose AI use, verify every output, protect data and keep a named human accountable for each conclusion. The same standard should apply to commissioners, who should also disclose how AI shapes their evidence mapping and commissioning decisions. Civil society networks should help design these governance rules and not only comply with them.
CONCLUSION
We hold that artificial intelligence must serve equity, justice and sustainable development and never replace human dignity. The six takeaways point the same way. What remains is to make them binding: in terms of reference, in budgets, in contracts and in who sits at the table. AHOA offers its network of more than 3,000 organisations across 113 countries, together with its experience in community engagement and M&E, to help turn these takeaways into practice.
Dr. Uzodinma Akujekwe Adirieje, FAHOA
CEO and Programmes Director, Afrihealth Optonet Association (AHOA)
2026 EvalforEarth Champion | President, African Refugees Council (ARC)
Abuja, Nigeria | afrepton@gmail.com | ceo@afrihealthcsos.org | www.afrihealthcsos.org
+234 80 347 25 905