During the first edition of the Impact Accelerator Program, the InfoNTD Consortium and OMP and Analytics for a Better World co-developed an AI chatbot to make technical information on neglected tropical diseases easier for frontline health workers and researchers to find and use. The first version focuses on cross-cutting NTD topics, with the longer-term ambition of extending the tool across all 21 neglected tropical diseases.
Why can finding the right NTD information be difficult?
The InfoNTD Consoritum brings together knowledge and tools on neglected tropical diseases, or NTDs — a group of diseases that affect more than one billion people worldwide and often occur in settings where access to specialist health information can be limited.
For frontline health workers and researchers, having information available does not automatically mean it is easy to use. Relevant guidance may be spread across many resources, and finding the right document or answer can take time.
The consortium wanted to make the information already available through InfoNTD easier to access, particularly for frontline health workers. This was especially important because this group had historically engaged less with the platform than other audiences, while they represent the first line of response.
The challenge was therefore not to create more information, but to make existing knowledge easier to find, understand and apply.
Why is this an AI and analytics problem?
A traditional search function requires users to know what terms to look for and then work through the results themselves.
A conversational AI tool can provide a different route. Users can ask a question in everyday languages such as French, Spanish, and Portuguese, receive a focused response and be directed back to the source material.
For the InfoNTD Consortium, the aim was not to replace expert guidance or the underlying knowledge base. It was to create a more accessible way into that information, as most sources are usually locked behind a paywall and written in English alone in academic terms.
The project also raised a broader data question: if searching and retrieving information can be made more efficient, could parts of the process for identifying, analysing and adding new resources to InfoNTD eventually be automated as well? The team also identified this as a possible future area of development from this project.
How have InfoNTD Consortium, OMP, and ABW developed the approach together?
The project was developed through the Analytics for a Better World Impact Accelerator Program, where the mission-driven consortium InfoNTD was paired with one of the industry leaders in optimisation, OMP.
The collaboration began with an in-person workshop in Rotterdam to clarify the problem, define what the tool should achieve and agree on how the team would work together.
Rather than moving immediately to one technical solution, the team developed two prototypes. These were then assessed through user interviews, an end-user survey and internal testing before one approach was selected for further development.
OMP contributed technical expertise, while the InfoNTD Consortium brought knowledge of NTD information needs, the existing platform and its users. The team also coordinated with Unc Inc, InfoNTD's existing technology partner, so that the chatbot could eventually fit into the organisation's wider technical environment.
Part of the InfoNTD Consortium with the OMP team at their first in-person workshop
How did users influence the design?
Reaching frontline health workers for interviews was one of the project's main challenges. Their time was limited, some were unfamiliar with InfoNTD, and unreliable internet connections sometimes made conversations difficult.
Rather than designing around assumptions, the team combined the interviews they could conduct with an end-user survey.
That feedback influenced several concrete design choices. The chatbot was designed to:
use simple and accessible language
offer multilingual support
ask follow-up questions when users need guidance
provide links back to the original information sources
The project also changed direction as evidence emerged. The partners did not follow a fixed development plan; they adapted the solution as prototypes, interviews and testing showed what users needed. These choices were intended to make the tool easier to use across different contexts rather than expecting users to adapt to a technically complex interface.
Why do sustainability and change management matter?
Building a chatbot was only useful if InfoNTD can maintain it after the accelerator programme ends. For that reason, the team considered operating costs and technical ownership during development rather than after the tool had been built.
Recurring technology costs could affect how widely and for how long the chatbot can be made available. That consideration influenced the decision to use a comparatively low-cost platform from the beginning. They selected Botpress, a low-code platform, so that the InfoNTD team can make basic changes themselves and the solution can be transferred more easily to Unc Inc for longer-term technical maintenance.
The collaboration itself also depended on change management. Meeting in person at the beginning, discussing expectations openly and holding regular weekly meetings allowed challenges to be identified early and decisions to be made jointly.
What can this mean for the InfoNTD Consortium next?
The current chatbot focuses on cross-cutting NTD issues. In the longer term, the InfoNTD Consortium aims to expand its knowledge base to cover resources and tools relating to all 21 neglected tropical diseases.
The chatbot's evolution will be shaped by two main considerations. The first is how new information is added. At present, expanding the chatbot's knowledge base is expected to involve a largely manual process. However, automating parts of the search, analysis and upload of new resources could become a new follow-up project.
The second is how the tool can reach a larger user group while keeping recurring costs manageable.
The next phase is therefore not simply about making the chatbot bigger. It is about testing whether frontline health workers find it useful, learning how they use it and developing a model that InfoNTD can maintain as its coverage grows.
The longer-term opportunity is to make the knowledge already available through InfoNTD easier to reach at the moment health professionals and researchers need it.
