AI in Education

Independent evaluation of an oral reading fluency model

Before a speech-based literacy model is deployed at scale, its accuracy and fairness need to be independently assessed. APLYD evaluated the system without involvement in its development.

Client / Partner Wadhwani AI

Sector Education; M&E / Responsible AI

Capabilities Independent assurance

Engagement Type Independent evaluation

Operational Context

Oral reading fluency tools use speech models to assess children’s reading performance. At scale, differences in model performance across learner groups can affect the reliability and fairness of the resulting assessments.

Institutional Delivery

We independently evaluated the model across approximately 8,000 Hindi and Gujarati recordings, assessing accuracy and bias across grades, dialects and gender groups before deployment.

Public Value & Lasting Impact

Independent evaluation provides an evidence base for deployment decisions by separating assessment from development. In this case, the model was assessed for accuracy and bias before wider use, providing evidence on its performance across different learner groups.

AI advisory in agriculture, Andhra Pradesh

Andhra Pradesh’s Rythu Seva Kendras are farmer service centres that support agricultural advice and services. APLYD advised the government on strengthening these centres as data-driven planning and advisory hubs, using AI and digital systems within existing public delivery structures.

Client / Partner Government of Andhra Pradesh

Sector Agriculture; Public Services / Government

Capabilities Public-service solutions; field evidence and service design

Engagement Type Advisory and system design

Operational Context

Rythu Seva Kendras provide a direct channel for agricultural services and farmer support. The objective was to strengthen these centres with better use of data and technology, while ensuring the approach could work across a large and diverse farming population.

Institutional Delivery

We advised on the design of data-driven planning hubs built around the existing centres, including advisory services, data pipelines and decision-support tools. The approach was designed to integrate AI with existing workflows and retain human oversight within service delivery.

Public Value & Lasting Impact

The work demonstrates how AI and data systems can be integrated into existing public-service infrastructure rather than developed as standalone applications. It combines digital decision support with established frontline delivery channels, creating a basis for more data-informed agricultural services at scale.

Farmer-centric data governance

Agricultural data is becoming increasingly important to how programmes are designed and delivered. That makes clear governance around data rights, consent and value-sharing essential. APLYD developed farmer-first frameworks to support this across agricultural programmes.

Client / Partner Gates Foundation

Sector Agriculture; M&E / Responsible AI

Capabilities Responsible adoption; governance

Engagement Type Governance framework

Operational Context

As agricultural programmes generate and use more data, institutions need clear frameworks for how that data is governed, including questions of rights, consent and value-sharing.

Institutional Delivery

We developed farmer-first data-governance frameworks that define how rights, consent and value-sharing should be addressed across agricultural programmes. The frameworks establish clearer principles for how agricultural data is governed and how the interests of farmers are considered within that process.

Public Value & Lasting Impact

Responsible use of agricultural data depends on the governance structures established around it. By addressing rights, consent and value-sharing at the programme level, institutions can create a stronger foundation for responsible data use and future AI applications.

Advancing AI adoption among India's MSMEs

Small manufacturers are an important part of India’s industrial base, but AI adoption requires a clear understanding of their readiness, constraints and capacity to implement. For the IndiaAI Mission, APLYD assessed where small manufacturers stood and what would support broader adoption.

Client / Partner MeitY, IndiaAI Mission

Sector MSMEs; Public Services / Government

Capabilities AI portfolio and readiness

Engagement Type Advisory and roadmap

Operational Context

The IndiaAI Mission needed a clearer understanding of AI-readiness across small manufacturers and the factors that influence their ability to adopt AI. The challenge was to move from broad ambition to an evidence-based view of what adoption would require in practice.

Institutional Delivery

We diagnosed AI-readiness across small manufacturers and developed a field-tested adoption roadmap for the IndiaAI Mission. The roadmap brought together practical playbooks, incentives and capacity-building plans, structured around the conditions and requirements identified through the assessment.

Public Value & Lasting Impact

Broadening AI adoption requires more than setting a national ambition. It requires an evidence-based understanding of readiness and a practical pathway for implementation. This work provided the IndiaAI Mission with a grounded basis for supporting AI adoption among MSMEs.

A shared AI sandbox for water utilities

AI applications for utilities need to be assessed in the context in which they will operate. APLYD supported the development of a shared environment where utilities across Africa could test AI applications against operational data before committing to procurement.

Client / Partner Africa Utility Data Collaborative

Sector Utilities; Public Services / Government

Capabilities Ecosystem infrastructure

Engagement Type Shared infrastructure

Operational Context

Utilities need evidence on how AI applications perform against their operational data and requirements before making procurement decisions. A shared testing environment can help institutions assess applications under relevant conditions before wider adoption.

Institutional Delivery

We developed a shared environment where utilities across Africa could test AI applications against operational data before committing to procurement. This created a common setting for utilities to assess applications and generate evidence to inform procurement decisions.

Public Value & Lasting Impact

Testing before procurement gives utilities a more informed basis for evaluating AI applications and comparing their suitability to operational requirements. A shared environment also creates an opportunity to build knowledge and evidence that can be relevant across multiple utilities.

A GenAI citizen services assistant for Kampala

The Kampala Capital City Authority wanted to extend access to city services through channels familiar to residents. APLYD developed a WhatsApp and Telegram assistant integrated with authority systems to support key service interactions.

Client / Partner Kampala Capital City Authority, Uganda

Sector Public Services / Government

Capabilities Public-service solutions

Engagement Type Product / deployment

Operational Context

Access to public services increasingly depends on how effectively digital channels connect citizens with government systems. The objective was to provide a more accessible interface while maintaining the security and system integration required for public-service transactions.

Institutional Delivery

We developed a WhatsApp and Telegram assistant covering registration, ticketing, document submission and status tracking. The assistant was integrated with authority systems through authenticated APIs, enabling service requests and updates to connect with existing government workflows.

Public Value & Lasting Impact

Effective citizen-facing AI depends on more than the conversational interface. It requires secure integration with the systems and processes behind the service. This approach combined a familiar digital channel with authenticated connections to government systems, supporting practical access to city services.

An AI evidence portal for a MEL programme

M&E programmes generate substantial volumes of evidence across reports, evaluations and field documents. APLYD developed a knowledge platform to make that evidence easier to retrieve, synthesise and use in programme decision-making.

Client / Partner Gates Foundation

Sector M&E / Responsible AI

Capabilities Operations and partner support; knowledge architecture

Engagement Type Product / knowledge platform

Operational Context

M&E programmes accumulate evidence across multiple reports, evaluations and programme documents. Bringing this evidence together to identify patterns, findings and lessons can be time-intensive, particularly when information is distributed across different formats and sources.

Institutional Delivery

We developed a knowledge platform with a domain-tuned model adapted to the programme’s terminology, methods and document structures. This enables teams to retrieve and synthesise relevant findings across programme documentation without reviewing every document in full.

Public Value & Lasting Impact

The usefulness of a knowledge system depends on how well it reflects the language, methods and context of the programme it serves. By adapting the model to that domain, the platform supports more efficient evidence synthesis and makes programme knowledge easier to access for decision-making.

AI across food and agriculture systems

AI in agrifood systems is evolving rapidly, but decisions on adoption need to be grounded in evidence about where it can create value and under what conditions. APLYD contributed to a mixed-methods study examining AI across agrifood systems in low- and middle-income countries.

Client / Partner 3ie and University of Birmingham

Sector M&E / Responsible AI; Agriculture

Capabilities Independent assurance; evidence review

Engagement Type Research study

Operational Context

AI applications are emerging across agrifood systems in low- and middle-income countries, creating a need for stronger evidence on their use, relevance and implications for smallholder-focused programmes.

Institutional Delivery

We contributed to a mixed-methods study of AI across agrifood systems in low- and middle-income countries, helping build an evidence base for responsible, pro-smallholder adoption.

Public Value & Lasting Impact

Responsible adoption requires a clearer understanding of where AI can contribute and the conditions required for it to deliver value. Evidence from cross-sector and country contexts can help governments, funders and other institutions make more informed decisions about the role of AI in agrifood systems.

AI in agriculture

AI for MSMEs

AI for utilities

AI for public services and government

Monitoring, evaluation and responsible AI

Independent evaluation of an oral reading fluency model

Operational Context

Institutional Delivery

Public Value & Lasting Impact

AI advisory in agriculture, Andhra Pradesh

Operational Context

Institutional Delivery

Public Value & Lasting Impact

Farmer-centric data governance

Operational Context

Institutional Delivery

Public Value & Lasting Impact

Advancing AI adoption among India's MSMEs

Operational Context

Institutional Delivery

Public Value & Lasting Impact

A shared AI sandbox for water utilities

Operational Context

Institutional Delivery

Public Value & Lasting Impact

A GenAI citizen services assistant for Kampala

Operational Context

Institutional Delivery

Public Value & Lasting Impact

An AI evidence portal for a MEL programme

Operational Context

Institutional Delivery

Public Value & Lasting Impact

AI across food and agriculture systems

Operational Context

Institutional Delivery

Public Value & Lasting Impact

Explore how APLYD can support your institution.