Terms of Reference: Consultant/firm to integrate an agentic feature into the health data governance legislative and regulatory assessment tool

Details of work required: Transform Health is seeking to recruit a qualified consultant/firm to design, develop, test and integrate an agentic feature into Transform Health’s Health Data Governance Legislative and Regulatory Assessment Tool (to automate the comparative analysis and generate the country summary).

Expected duration: the work should be carried out between October 2026 and February 2027 (5 months)

Background

Transform Health has developed tools to support governments and other stakeholders to strengthen health data governance frameworks. These include the Health Data Governance Principles which provide a shared, equity- and rights-based foundation for the governance of health data, and the Health Data Governance Legislative Blueprint (Model Law), which provides guidance for strengthening national legal and regulatory frameworks.

Transform Health has also developed a National Legislative and Regulatory Assessment Tool, which supports a document-based review of the national health data governance landscape. The tool provides a structured three-step approach for reviewing national legislation and regulations, against the Health Data Governance Principles. This includes: 1)  identifying and categorising national legislation and regulatory documents to be assessed; 2)  mapping and analysing how the Health Data Governance Principles are reflected in the selected legislative and regulatory documents; and 3) creating a summary of the analysis. As steps 2 and 3 of the process are completed manually (including reviewing legal documents, populating worksheets, and summarising the analysis) the aim of this consultancy is to help automate the process using AI. 

The assessment framework already exists. This consultancy translates it into a reliable agentic workflow that processes uploaded documents and produces a populated draft assessment for human verification. Given the accuracy and traceability required in legislative assessment, the feature must be designed as an evidence extraction tool, not a system that interprets legislation or generates legal findings. The tool is expected to be used at low and intermittent volume, likely a small number of country assessments per month. The system should therefore be designed and optimised for cost-efficiency, reliability and ease of maintenance at low volume,The assessment tool sits on the Health Data Governance website, which is built on WordPress. The new feature will need to be appropriately integrated into this environment on the Health Data Governance website.

Scope of work

There will be regular project delivery check-ins to monitor progress against the indicated timelines for the deliverables. 

1. Orientation and initial technical design

• • Review the Health Data Governance Legislative Assessment Tool, including existing framework for analysis and steps,  assessment worksheets, and the technical environment of the existing Health Data Governance website.
  • Reviewing the current hosting environment, assessing whether it can support the document processing required.
  • Define the functional and technical requirements for the agentic feature, including:
    • Requirements for user registration and access to the tool, including the information collected at sign-up, verification of registrations, and any approval step
    • Define acceptable and standardised input formats 
    • Step 2 and Step 3 output formats
    • Requirements for user consent regarding the storage, use and potential publication of completed Step 2 and Step 3 outputs
    • User workflow from document upload to download of completed worksheets;
    • Requirements for document processing, citations, job status and error handling;
    • Data handling, security, retention and deletion requirements for uploaded documents and generated outputs (taking into consideration data privacy requirements and compliance)
    • Technical architecture and any additional infrastructure, APIs or third-party services required
    • An assessment of whether the existing hosting environment can support the required document processing, and a recommendation on any additional infrastructure
    • Cost controls, including per-assessment limits on the number of documents and total pages, per-job token caps, and any usage quotas per registered user
    • Confirmation that the architecture is model- and provider-agnostic, so that the underlying AI model or provider can be changed without redesign.

2. Design and develop the agentic system 

• • Develop the agentic system (backend rules and mechanisms) to process uploaded legislative and regulatory documents in three languages (English, French, Spanish), run the comparative analysis against the Core Elements of the Health Data Governance Principles and their standard interpretations, as well as the Legislative Blueprint, and identify and extract relevant text from the uploaded documents to populate a worksheet (with verbatim text and clear citation to the source document), as outlined in Step 2 of the Tool.
  • Develop safeguards to minimise hallucination and unsupported outputs, including ensuring that the system does not invent, paraphrase or generate legislative provisions or present unsupported interpretations as findings. 
  • Develop the functionality to automatically generate a summary view of the detailed analysis (step 3 of the tool),as well as a summary report of findings, from the step 2 results, using transparent, pre-defined and deterministic rules. 
  • Develop appropriate document parsing functionality, including a defined failure pathway for scanned or non-machine-readable documents.
  • Optimise the system for token efficiency without compromising accuracy, including the use of prompt caching, batch processing where the asynchronous workflow allows, and an approach to document retrieval that balances cost against the risk of missing relevant text. The proposal and design specification should set out this trade-off explicitly and how recall will be measured.
  • Implement a configurable per-job token and cost cap, with graceful termination and a clear message to the user if an assessment exceeds it.

3. Build and integrate the user interface

• • Develop a user-friendly interface that enables users to:
    • Upload the completed Step 1 worksheet and relevant legislative and regulatory documents;
    • Initiate an assessment and view processing status;
    • Receive notifications where documents cannot be processed successfully; 
    • Download the completed draft Step 2 and Step 3 worksheets;
    • Download the summary report of the analysis.
  • Develop a registration mechanism through which users request access to the tool before running an assessment.
  • Implement the processing functionality in the environment agreed at design stage, with the Health Data Governance website providing the user interface, upload, status display and download functions.
  • Ensure that the initial version of the feature supports English, French and Spanish inputs and outputs.

4. Accuracy testing and user acceptance testing

• • Test the functionality and outputs against an agreed sample of existing country legislative assessments to verify accuracy and address any hallucination issues.
  • Develop and apply a testing methodology and pre-agreed performance thresholds to assess the accuracy, completeness and traceability of extracted findings, and to test the registration, verification, acknowledgement and consent flows.
  • Identify and address false positives, missed relevant text, citation errors and any hallucination or unsupported output issues.
  • Conduct user acceptance testing with a small group of pilot users and/or countries to assess usability, clarity and the quality of outputs.
  • Refine the functionality based on technical testing and user feedback.
  • Measure and report actual token consumption, cost per assessment and processing time across the test sample, at both typical and maximum corpus size.

5. Deployment, launch and handover

• • Deploy and launch the final tool on the Health Data Governance website.
  • Develop clear user guidance explaining the purpose of the feature, required inputs, how to use the tool, the Step 2 and Step 3 outputs, the meaning of “Not identified”, and the importance of human review and verification.
  • Prepare technical documentation and handover materials, including recommendations for ongoing maintenance, monitoring and future development.

Key deliverables and deadlines

Profile and experience

We are looking for a consultant/firm with demonstrated expertise in developing agentic systems and integrating them into existing interfaces document analysis and software development, with the ability to develop reliable, evidence-grounded tools for use by non-technical users.

The successful consultant/firm should demonstrate:

• • Expertise in AI and document intelligence, including natural language processing, large language models, information retrieval and the analysis of complex or lengthy documents.
  • Experience developing evidence-grounded agentic systems, with demonstrated approaches to minimising hallucination, ensuring outputs are traceable to source documents, and providing accurate citations.
  • Strong software development and technical architecture experience, including the development of secure, reliable and scalable web-based applications and AI-enabled tools.
  • Experience in document processing and structured data extraction, including processing documents and developing appropriate approaches for scanned or non-machine-readable documents.
  • Experience integrating digital tools into websites or content management systems, preferably including WordPress, and familiarity with cloud infrastructure, APIs, data security and the handling of uploaded user documents.
  • Experience testing and refining AI-enabled tools, including benchmarking against existing datasets or assessments, defining performance measures, conducting user acceptance testing and translating technical findings into practical improvements.

Desirable: Experience working with legislative, regulatory or policy documents, and/or in health, digital health, or health data governance would be an advantage.

Remuneration

The proposed budget for this assignment is approximately USD 15,000 (to be agreed based on experience and the proposal).

Application process 

To apply, please submit the following to hr@transformhealthcoalition.org by 09:00 CET on 21 September 2026: 

• • Your CV and cover letter outlining your experience relevant to this consultancy.
  • A brief proposal outlining your approach to this assignment.
  • 2-3 samples of similar work.
  • Your proposed quote, including a breakdown by key tasks/deliverables.

Transform Health is a global coalition of more than 230 organisations dedicated to achieving universal health coverage (UHC) by 2030 by harnessing digital technology and the use of data. We bring partners and networks together, leveraging our shared experience and expertise, to shape policy agendas, influence decision makers, develop tools, provide technical support to governments and others, and conduct research and analysis to deliver health for all in the digital age.