The model is evidence-aware, but each engagement must still prove what works in context.

Evidence posture

GI is built on a clear thesis: public systems often fail not because nobody knows anything, but because learning does not travel, coordination is weak, evidence is hard to use and solutions are not adapted to context.

Starting assumptions

• Implementation and coordination capacity shape whether policy intent becomes public value.
• Embedded actors generate practical knowledge that does not always travel.
• Adaptive problem-solving needs diagnosis, experimentation, review and learning capture.
• Knowledge infrastructure should be curated and honest about adoption conditions.
• Capacity building is stronger when it is tied to live governance work and role expectations.

Evidence hierarchy

The note separates three levels of evidence: public literature that supports the design logic, operating hypotheses that GI must test in practice and validated field results that should only be claimed after review.

• Design support: public references can justify why the model is plausible.
• Operating hypothesis: each engagement should state what GI expects to learn or improve.
• Validated result: public outcome claims need reviewed evidence, context and limits.

What each engagement must validate

A state engagement should test whether embedded governance talent can anchor structured Action Labs, whether review cadence improves continuity, whether curated pathways reduce repeated reinvention and whether GovCap products make field learning reusable for new actors.

Validation questions

• What public problem was specific enough to diagnose, review and adapt around?
• What changed in routines, coordination, evidence use or decision quality?
• What did not work, and what conditions limited adoption or durability?
• What knowledge object, role guide or learning product can be reused without exposing sensitive detail?

Reference base

The public evidence frame draws from work on governance and implementation, service-delivery accountability, adaptive problem-solving, Doing Development Differently and evidence-use practice. These sources support the design assumptions; they do not prove GI-specific outcomes.

• World Development Report 2017: Governance and the Law
• World Development Report 2004: Making Services Work for Poor People
• Problem Driven Iterative Adaptation
• Doing Development Differently Manifesto
• Alliance for Useful Evidence

Learning products

The output should not only be a report. Useful evidence becomes playbooks, failure notes, solution pathways, evaluation questions, role-specific modules and Collective entries that other governance actors can use with appropriate context.

Public claim discipline

GI should not overstate proof before an engagement is operating. Public claims should distinguish between broad evidence behind the approach, early operating hypotheses and validated field results.

This public evidence note is not an evaluation report. It is a launch-stage framing document for conversations with governments, funders, practitioners and knowledge partners.