What is APLYD's delivery framework for applied AI?

APLYD follows a structured four-part journey: 1) Assessment & Strategy (evaluating readiness, priorities, and workflows), 2) Solutions & Architecture (designing around public service realities), 3) Field Evidence & Service Design (studying frontline delivery contexts), and 4) Institutional Capability Transfer (handing over documentation, code, and operational routines).

How does APLYD ensure AI solutions fit existing workflows?

Interoperability, security, and decision rights are designed directly into the system architecture alongside frontline teams, connecting with existing MIS and operational systems rather than requiring disruptive IT overhauls.

Why is independent assurance separate from AI delivery?

To guarantee objective, unconflicted evidence. APLYD only evaluates systems that no one in its delivery group built, and does not bid to build systems it has evaluated.

From the decision worth improving to a system that lasts

Applied AI is easier to pilot than to sustain. Whether it creates lasting value depends on the decisions made before implementation: the problem being addressed, the evidence behind it, the conditions for deployment, and the institutional capacity to own and operate what is built.

Explore

Strategy and direction

The first stage is about establishing where AI can create meaningful value and what it will take to deliver it. Rather than beginning with a technology or a predetermined solution, we start with the decision, service or workflow that needs to improve, and the people and institutions it affects.

This means assessing the mandate, availability and quality of data, existing infrastructure, operational workflows and institutional capacity required for implementation. We examine where AI can improve decision-making or service delivery, and where it may introduce unnecessary cost, risk or operational complexity.

Readiness and risk are assessed together. A promising use case is only viable when the underlying data, institutional mandate, operating environment and ownership arrangements can support it. The outcome of this stage is a clear view of the opportunity, the conditions required for implementation, and whether the use case is ready to move forward.

Strategy and direction consultation and institutional governance

Design

Field and delivery

Good design begins with the delivery setting. Before deciding what to build, we understand how work actually happens, where systems fall short, and what people do to bridge those gaps.

We observe frontline workflows, trace cases through the system, and work alongside the teams responsible for delivery. This brings forward the last-mile evidence that is often lost between frontline experience and institutional decision-making, while making the real constraints of implementation visible.

We then co-design and prototype with the people who will use the system, testing early versions against real workflows rather than assumed ones. The result is not simply a better-designed tool, but one shaped by the context in which it needs to work. At the end of this stage, the idea is either refined into something worth building or stopped before further resources are committed.

Deliver

Design, build and scale

The third stage turns a validated prototype into a system that can operate reliably within an institution’s existing environment. This requires aligning the product, architecture, data, workflows and governance arrangements needed for sustained use.

We undertake the AI and data engineering, integrate the system with existing pipelines and operational systems, and define requirements for interoperability, security, access and decision rights as part of the design.

Rollout is treated as part of implementation. We establish the training, documentation, support and operating structures needed for institutional teams to manage the system from the outset. The aim is not simply to deploy a working system, but to build the capability and ownership required to sustain it over time. Scale follows evidence and operational readiness.

Systems engineering, pipeline integration and institutional handover

Assure

Evaluation and assurance

The final stage assesses whether a system is performing as intended and delivering value for the people and institutions it was designed to serve.

We assess model performance, user adoption and workflow outcomes, alongside safety, fairness and broader impact. The findings inform a clear decision: improve, hold or scale. Where the evidence does not support continued use, stopping is a valid outcome.

Evaluation is kept independent from delivery, including from our own work. We do not evaluate systems built by our group, and we do not bid to build systems we have evaluated. This separation allows institutions to assess performance, risks and outcomes without a conflict of interest.

The objective is to ensure that decisions about improvement and scale are based on evidence, not simply on whether a system has been deployed.

Independent empirical evaluation, fairness and algorithmic auditing