How AI changed my work as an evidence intermediary
Author: Laura Boeira1,2
1. Instituto Veredas
2. Latin American and the Caribbean Evidence Hub (HubLAC)
[I should warn readers: This is the tale of a VERY sceptical AI user.]
I became an evidence intermediary in 2016, when I resigned from my civil service position at the Ministry of Health and began work on creating Instituto Veredas, a Brazilian non-governmental organisation raising awareness of the importance of evidence-informed policymaking. An evidence intermediary (some people might prefer ‘evidence broker’ or ‘knowledge translator’), to give my personal definition, is someone who tries their best to facilitate communication between evidence producers (usually researchers, but sometimes members of civil society organisations or advocacy groups) and evidence users (i.e. decision-makers, who may be citizens or staff of governments and health services). I entered into this work because I really want policies and interventions to be informed by the best available evidence in a timely manner. In Brazil, this usually requires not only translating information from various languages into Portuguese but also summarising scientific knowledge into accessible forms, such as evidence briefs, rapid reviews, and even simple slide decks.
It seems complex, but it is really human: we speak with people who have important decisions to make and need urgent information, we help them frame their information needs as questions, and then we either connect with research teams or rigorously search the global evidence, incorporating any local evidence that is available. Then, we synthesise all of this information in an output that is ready to use, while transparently reporting where we lack certainty. And then we follow this cycle again and again – it seems there is no end to the stream of urgent questions in Brazil.
We often work with government branches, and we have valuable partners in the Brazilian Coalition for Evidence and the Brazilian Evidence-Informed Policy Network, who help us standardise methods and guarantee collaboration. Yet, while this long introduction may give the impression that everything works perfectly, there have been plenty of challenges in fostering evidence use in our country. For instance, one of our partners, the Applied Economics Research Institute (Ipea), surveyed civil servants in 2019 and found that most made policy-related decisions through consultation with colleagues (75.9%) or based on their own experience (64.6%), rather than referring to scientific studies (19.8–30.5%, depending on document type).
[You might be thinking, Laura, you are halfway into this blog and still haven’t mentioned AI! It is coming, I swear.]
For a long time, I believed our greatest challenge was to encourage all those civil servants to value evidence as a useful piece of the puzzle. But, cut to 2022, and while my team was away on a strategic planning retreat, a co-director remarked that, ‘In five years, our work will be basically extinct because of AI.’ I laughed so hard at the time. In 2022, I could not imagine a world where AI would be able to do all we did – and, even if it did, I was sure you would still need human mediation of its use to guarantee we were being explicit about its limitations. Less than one year later, however, in my hometown, a legislator created a policy bill solely using AI, didn’t disclose this to his colleagues, and the bill was approved. Then, within three years of my co-director’s premonition, our Federal Government published a guide on developing prompts that allow civil servants to conduct research using AI. It was then that I stopped laughing.
Today, many of us – whether policymakers, health workers, or citizens – have so many decisions to make every day that we appreciate AI’s capacity to provide opportune answers that are reliable enough, in whatever language or format we need. Yet, in my line of work, AI tools still have many biases, hallucinations, and limitations that stand in the way of their capacity to conduct trustworthy evidence synthesis. We are one year from my colleague’s prediction that the technology will have put us out of our jobs, and while I don’t see my evidence intermediary work being lost, AI has transformed every evidence ecosystem with which I interact.
Nonetheless, I am optimistic about the future. And why do I remain hopeful, you might ask? Well, because my brilliant colleagues around the world are already demonstrating how evidence intermediaries can adapt and collaborate with AI tools to be more effective:
Alongside its national government in South Africa, the Pan African Collective for Evidence (PACE) has developed a chatbot connected to an evidence map on gender-based violence and femicide. This allows policymakers to access a curated evidence bank and interact with its information, guaranteeing speed and trust at the same time
The Center for Rapid Evidence Synthesis (Acres), in Uganda, has launched a platform called the Living EIDM ToolMap, a collaborative repository of digital tools to assist policymakers and knowledge brokers in facilitating evidence-informed decision-making
The Center for Interinstitutional Collaboration on Artificial Intelligence Applied to Public Policies (CIAP), in Brazil, has created a Research Network on Artificial Intelligence for the Improvement of Public Policies, while the Artificial Intelligence and Health Center for Latin America and the Caribbean (CLIAS) has established a knowledge community to support the region.
For my fellow evidence intermediaries around the world, my key message is that we have many roles to play in this transforming ecosystem. For instance, you can choose if you want to develop or test new tools; contribute to standards that help guide decision-making; strengthen relationships between the communities of evidence users and producers, so they are not only mediated by AI; or support contextualisation and equity considerations through participatory processes that help include citizen voices and guarantee the democratic appraisal of evidence.
Let’s not duplicate our efforts – instead, we should collaborate in determining where AI can provide assistance.
At the same time, let’s never lean into overpromising – our best allies are still transparency and uncertainty.
NB. Because I am only human, I still rejoice when my handmade search string produces better results than an AI tool, or when I find information faster (and more accurately) with a Google search than my colleague who asks ChatGPT. But even a sceptic can admit that creating images is way more fun now AI has entered the room.
This text was fully written by a Portuguese speaker making a point to write in English to overcome the need to use AI for translation. The landscape image above was created by ChatGPT.
References
Global Commission on Evidence to Address Societal Challenges (2022). The Evidence Commission report: A wake-up call and path forward for decision-makers, evidence intermediaries, and impact-oriented evidence producers. McMaster Health Forum. https://www.mcmasterforum.org/docs/default-source/evidence-commission/evidence-commission-report.pdf?Status=Master&sfvrsn=2fb92517_5/Evidence-Commission-report
Haby, M. M., Chapman, E., Clark, R., Barreto, J., Reveiz, L., & Lavis, J. N. (2016). What are the best methodologies for rapid reviews of the research evidence for evidence-informed decision making in health policy and practice: a rapid review. Health Research Policy and Systems, 14, 83. https://doi.org/10.1186/s12961-016-0155-7
Koga, N. M., Palotti, P. L. D. M., Couto, B. G. D., Nascimento, M. I. B. D., & Lins, R. D. S. (2020). O que informa as políticas públicas: Survey sobre o uso e o não uso de evidências pela Burocracia Federal Brasileira [What informs public policies: Survey on the use and non-use of evidence by the Brazilian Federal Bureaucracy] (TD 2619). Instituto de Pesquisa Econômica Aplicada [Institute of Applied Economic Research] (Ipea). https://doi.org/10.38116/td2619
Ministério da Gestão e da Inovação em Serviços Públicos [Ministry of Management and Innovation in Public Services] (2025). Guia prático de prompt e pesquisa com IA para servidores públicos [Practical guide to AI prompt and search for public servants]. https://www.gov.br/governodigital/pt-br/infraestrutura-nacional-de-dados/inteligencia-artificial-1/publicacoes/guia-pratico-de-prompt-e-pesquisa-com-ia-para-servidores-publicos
Sousa, M. S. A., Peiris, S., Figueiró, M. F., Haby, M. M., Baraldi, A. C., Reveiz, L., & Souza, J. P. (2026). The landscape of artificial intelligence tools and platforms for evidence synthesis: A scoping review. Systematic Reviews, 15, 82. https://doi.org/10.1186/s13643-025-02842-y
To link to this article - DOI: https://doi.org/10.70253/OKPH8887
Conflict of interest
Laura is a member of the World EBHC Day Steering Committee.
Disclaimer
The views expressed in this World EBHC Day Blog, as well as any errors or omissions, are the sole responsibility of the author and do not represent the views of the World EBHC Day Steering Committee, Official Partners or Sponsors; nor does it imply endorsement by the aforementioned parties.
