AI Assisted CLM Analysis and Lightweight Dashboard Development

1. Introduction:

The International Treatment Preparedness Coalition Global (ITPC), is a global network of community-led organizations working to improve equitable access to treatment and health services. ITPC partners with civil society, governments, and technical agencies across more than 60 countries to strengthen community leadership, accountability, and data use within health systems.

Through community-led monitoring, research, and technical assistance, ITPC supports the integration of community-generated data into decision-making and service improvement. This Request for Proposals seeks a qualified partner to support ITPC and country partners in strengthening analysis, reporting, and practical use of community data for timely action.

2. Background and Context

ITPC is responding to the global emergency in HIV service provision resulting from funding reductions and policy shifts affecting key populations. Through our Community-Led Monitoring (CLM) initiatives, we collect critical data from recipients of care and service providers worldwide to document healthcare access barriers, service disruptions, and rights violations. Our data collection spans routine monitoring and time-sensitive contexts requiring rapid validation and response.

As part of our expanded monitoring mandate, we seek to enhance our analytical capabilities to process both routine CLM data and emergency signals, creating a public-facing visualization system that communicates both validated trends and identified priority issues while preserving respondent anonymity. This RFP solicits proposals from technical partners who can help us implement AI-powered tools to accelerate analysis of mixed-method data and develop an interactive dashboard system that serves monitoring, validation and reporting functions.

Current processes in South Africa involve extensive manual cleaning, coding, integration, and drafting. Manual qualitative coding can take 3 to 4 weeks in Malawi. The system must reduce this significantly while preserving expert oversight and data validation. The dashboard must support facility-level summaries and basic geographic drill-down (facility to district to national levels).

3. Purpose

ITPC seeks a technical partner to design and implement a practical, AI-assisted analysis workflow and lightweight dashboard using primarily configurable, off-the-shelf tools for CLM across South Africa, Malawi and potentially other countries in the region.

The system must reduce manual analysis time, improve data quality, and generate clear outputs for advocacy and decision-making. The scope is limited to a maximum allocated project budget inclusive of development, training, documentation, hosting and first year support.

4. Core Objectives

The following core objectives must contribute to reducing the analysis-to-action cycle, while preserving community oversight and validation authority:

• Develop AI-assisted qualitative synthesis and quantitative trend summaries
• Enable narrative clustering and urgency classification
• Generate automated draft briefs and monthly summaries
• Provide translation support for key local languages where feasible
• Design a lightweight internal review interface
• Deliver a lean public-facing dashboard

The solution is expected to rely on existing AI APIs and business intelligence platforms rather than custom-built machine learning models.

5. Foundational Design Principles

The proposed solution must adhere to the following design principles:

1. 1. Community-Led Governance: All AI outputs must remain advisory. Final interpretation, validation, and release decisions remain under community and staff oversight.
   2. No Fully Automated Public Alerts: Emergency signals must pass through human validation workflows before public dissemination.
   3. Augmentation, Not Replacement: AI tools must accelerate synthesis and pattern recognition but must not substitute for community-defined indicators or decision-making processes.
   4. Context Preservation: Qualitative narratives must not be reduced to sentiment scores without preserving contextual meaning.
   5. Stakeholder Acceptability: The systems must foster trust and acceptance among public health stakeholders.

6. Design Context

The proposed system must operate effectively in environments where:

• • Trust in institutions may be low
  • Data may be incomplete or politically sensitive
  • Community authorization is essential for data validity

The proposed system/platform must therefore assume community-led oversight as a structural requirement.

7. Scope of Work

PART A. AI-Assisted Analysis System

The vendor will develop an AI-assisted analytical layer focused on:

A. Data Processing and Cleaning

• Flag potential outliers using predefined statistical rules (e.g., threshold or deviation-based)
• Identify duplicates
• Highlight missing values
• Display error indicators clearly for user review
• Generate automated descriptive trend summaries (month-on-month, quarter-on-quarter)
• Allow disaggregation of data (age, sex, key population group)
• Flag partial/incomplete responses

B. Quantitative Trend Summaries

• Generate automated indicator trend tables
• Provide aggregation at facility, district, and national levels
• Enable rule-based anomaly flagging using configurable thresholds (no predictive modeling required)
• Ensure exportable charts and tables
• Allow group comparisons

C. Qualitative Synthesis

• Enable thematic clustering of transcripts and open-text responses
• Provide keyword extraction
• Apply urgency tagging based on configurable categories
• Auto-generate draft narrative summaries

AI outputs must remain advisory. Human validation is required before publication. The system may utilize third-party AI APIs for text analysis and summarization. Custom model training is not required.

D. Translation Support

• Ensure translation of selected local languages to English
• Leverage commercial AI translation to deliver reasonable translation accuracy. No custom language model development is required
• Provide clear indications when translation confidence is low

E. Automated Draft Outputs

• Monthly summary briefs
• Facility-level snapshot summaries
• Quarterly synthesis report drafts

These outputs must be editable before release. Templates may be predefined and configurable rather than dynamically generated from scratch.

PART B. Lightweight Internal Review Interface

The system must include:

• Secure login
• Role-based access
• Data validation workflow
• Draft report review interface
• Audit logs

The interface must be usable by non-technical staff and reduce reliance on consultants.

PART C. Lean Public-Facing Dashboard

The public dashboard must:

• Display validated indicators only
• Allow geographic drill-down
• Provide optional facility comparison views (e.g., quartile grouping) where appropriate
• Be mobile responsive
• Function in low-bandwidth environments

The dashboard must not include fully automated emergency alerts, and the architecture must remain adaptable for future expansion.

8. Technical Requirements

8.1 Integration Requirements

8.1.1 The system must:

• Import Kobo Toolbox, Alchemer and CommCare data via API or structured export (CSV/Excel acceptable)
• Export structured CSV compatible with DHIS2
• Allow configurable refresh schedules

8.1.2 Measurement of impact

Vendors must propose clear metrics to demonstrate improvement. At minimum, the system must measure:

• Reduction in time to clean datasets
• Reduction in time to produce trend analyses
• Reduction in time to code transcripts
• Reduction in total turnaround time from collection to dashboard updates
• Error rate detection improvements

These measures reflect partner-defined efficiency and quality indicators. Measurement will focus on efficiency gains rather than predictive accuracy.

8.1.3 Vendors must propose a baseline and projected improvement target.

8.2 Dashboard Specifications:

• Reliable commercial cloud hosting with standard availability appropriate for NGO-scale usage
• Progressive enhancement design ensuring core functionality works in low-bandwidth environments
• Scheduled data refresh (daily or weekly configurable). Near-real-time processing is not required
• Role-based access controls with granular permissions for data access and management
• Audit logging of data uploads, validation actions and report approvals

9. Budget Envelope

Maximum contract value is within the allocated project budget and must cover:

• Development
• Configuration
• Training
• Documentation
• Hosting for 12 months
• Support during pilot

Competitive bids are encouraged. Preference will be given to proposals that demonstrate efficient use of existing commercial tools and limit custom development. Proposals that exceed the available budget will not be considered.

10. Sustainability Requirements

The system must:

• Require minimal coding for ongoing updates
• Include user manual and training materials
• Allow full data export
• Avoid vendor lock-in

Training must be provided for local teams. Preference will be given to modular, low-code or no-code components where feasible.

11. Deliverables Timeline

• Deliverable 1: Requirements analysis and system design

Projected Timelines: April 2026

Advocacy & Operational Use Approach: Finalize indicator definitions, rule-based flag thresholds, validation workflows, reporting templates

• Deliverable 2: AI-assisted analysis configuration

Projected Timelines: April – May 2026

Advocacy & Operational Use Approach: Configure qualitative summarization, urgency tagging, translation services, and quantitative trend summaries using commercial AI APIs

• Deliverable 3: Dashboard configuration and testing

Projected Timelines: May – June 2026

Advocacy & Operational Use Approach: Configure lightweight public and internal dashboards using business intelligence tools, implement geographic drill-down and indicator filtering

• Deliverable 4: Data integration and staff training

Projected Timelines: June – July 2026

Advocacy & Operational Use Approach: Implement API or structured data imports, train staff on AI-assisted review, validation workflows, and dashboard use

• Deliverable 5: Pilot implementation in 2 regions

Projected Timelines: July – August 2026

Advocacy & Operational Use Approach: Test analysis workflow with real CLM data, refine summaries, thresholds and templates based on user feedback

• Deliverable 6: Launch and 12-month support

Projected Timelines: August 2026

Advocacy & Operational Use Approach: Deploy production system; provide support, minor refinements, and monitoring of efficiency improvements

12. Proposal Submission Requirements

Interested technical partners should submit proposals including:

1. 1. Organizational Profile: Company background, relevant experience with health/NGO data, health monitoring and data visualization systems and team qualifications.
   2. Technical Approach: Detailed methodology for achieving AI-analysis, dashboard, and threshold-based alert functionality.
   3. Integration Strategy: Dedicated subsection explaining methodology for connecting with Alchemer, Kobo Collect, DHIS2, and ability to adapt configuration for new indicators if required.
   4. Advocacy & Emergency Response Approach: Explanation of how tools will shorten analysis-to-action cycle for both routine advocacy and operational readiness for routine monitoring use.
   5. Portfolio Examples: Similar projects completed, especially those involving health monitoring, early warning systems, or emergency response platforms.
   6. Implementation Plan: Timeline with milestones accounting for pilot testing and iterative refinement of the systems.
   7. Budget Breakdown: Detailed costing including development, training, maintenance, hosting and post-implementation support provisions.
   8. Support Model: Post-implementation support with specific attention to systems reliability and operational continuity.
   9. References: Two client references from similar projects, ideally including health monitoring implementations.

j. Ethical Framework: Proposed approach to ethical challenges in AI-driven data handling

and processing.

13. Evaluation Criteria

Proposals will be assessed on:

• Practical feasibility within project budget ceiling
• Demonstrated experience with health data platforms
• Clear plan to reduce analysis time
• Data quality and governance safeguards
• Simplicity and sustainability
• Ability to deliver within timeline

Proposals that rely on custom AI model development will receive a lower score than those leveraging configurable, commercially available AI services.

English