Civic intelligence infrastructure

There are thousands of civic initiatives — and they mostly operate in disconnected silos, precisely when coordinated responses are most needed. Civic intelligence infrastructure (CII), a concept developed by Brandon Nørgaard for the Mediators Foundation with Better Together America, is a proposed answer at two different scales: the coordination layer that lets organizations and their AI agents work together across affiliate networks, and the local community infrastructure that grounds it.

The problems it names

• Invisibility — groups doing complementary work never discover each other, even a few miles apart. A civic hub running effective bridging dialogues in one region has no way to find a facilitator network elsewhere that has already built the training materials and conflict-navigation tools it needs — so both keep reinventing the same resources independently.
• Coordination failure — no easy way to find collaborators across geographic or functional boundaries, even when organizations working on complementary pieces of the same problem operate in the same area without ever connecting.
• Fragmented infrastructure — every organization runs separate event, CRM, and measurement tools that can’t talk to each other.
• Infrastructure dependency — community and civic operations run on infrastructure owned and governed by outside parties, with little local control.
• Agent coordination gap — as organizations adopt AI agents internally, those agents have no shared way to communicate or coordinate with agents at partner organizations or affiliate sites — across different vocabularies, formats, and assumptions.
• Impact opacity — communities know their work is changing things — people feel more confident evaluating information, relationships are forming across old divides — but have no way to quantify those shifts or connect them to the longer-term outcomes funders and communities actually care about.
• Knowledge fragmentation — each hub reinvents what another already solved. An innovative civic process developed in one community stays unknown to hundreds of others facing the same disengagement challenges.
• Systems blindness — groups treat symptoms without seeing the system they’re part of, working in parallel on different facets of a shared root cause without ever connecting to each other.

The idea: two scales, two layers

CII proposes two layers. A normalization layer translates the data from each group’s existing tools into a shared “civic data taxonomy” — so coordinators keep the tools they already use, and a common standard runs underneath. On top of that sits a backend intelligence layer: ecosystem mapping (organisations, relationships, resource flows, gaps) and aggregate analytics that make scattered local action legible as one coordinated movement — for network coordinators, funders, and researchers. It’s meant to integrate existing civic tech, not replace it.

Layer 1 addresses the networked scale: how NGOs with local affiliate networks — and the multiple AI agents those organizations increasingly run, both internally and across organizational boundaries — can communicate and coordinate. This is the connective tissue and intelligence layer. It works at the level of translation, grammar, and semantics — giving agents and organizations that use different vocabularies, data formats, and internal logics a shared basis for interoperability and interactability. On top of that, it normalizes activity and data from across affiliates and partner organizations into a shared structure, enabling ecosystem mapping and aggregate analytics that make scattered local action — and scattered agent activity — legible as one coordinated effort, for network coordinators, funders, and researchers. The aim is to make visible the kind of impact that’s normally invisible: not just event counts, but the relationships, trust, and systemic shifts that communities are actually trying to create.

Layer 2 addresses the local scale: how to establish community-controlled, locally-governed mini data centers, and what bundle of tools to provision them with to support community, civic, and guild operations. Rather than civic groups depending on infrastructure owned and operated by outside platforms, Layer 2 is infrastructure a community can run and govern itself — physical hardware, hosted locally, paired with a practical software stack: event management, community relationship tracking, asset mapping, deliberation tools, and lightweight measurement, packaged for teams without dedicated technical staff. The thinking here connects to work on verifiable trust communities built from the ground up — establishing the kind of locally-rooted, accountable infrastructure that a community can trust because it controls it. The goal at this scale is direct and practical: less administrative burden for the people doing the work, so more of their time goes to the community itself rather than to wrangling spreadsheets and disconnected tools.

These two layers operate at different scales and can each advance on their own. But they’re complementary: the local infrastructure in Layer 2 is a natural source of the data and activity that Layer 1 helps coordinate and make visible at the network level, while insight generated at the Layer 1 scale — what’s working elsewhere, where resources should flow, which communities are facing similar challenges and could learn from each other — can flow back down to inform local Layer 2 operations. Interoperability between the two is a goal, though not necessarily an immediate priority.

Why not just rely on AI?

It’s tempting to think a sufficiently capable AI could just absorb all this civic data and produce the coordination and insight CII is after. But civic intelligence — used by funders, network coordinators, and communities themselves to make real decisions — needs a level of rigor that a pure language-model approach can’t guarantee on its own. Language models can hallucinate, can produce different analyses from the same data on different runs, and don’t inherently provide a way to trace a conclusion back to the evidence behind it. For communities whose work is chronically undercounted and underfunded, an opaque or inconsistent analysis is worse than no analysis at all — it risks making their impact look smaller, or different, than it actually is.

CII’s design instead keeps a knowledge graph as the source of truth — defining what counts as valid and tracing every claim back to the instrument and organization it came from — and uses language models only at the edges (natural-language data entry, drafting reports, surfacing patterns). The goal is provenance and consistency over a confident black box. The same logic applies to agent coordination at the Layer 1 scale: agents from different organizations need a shared, structured basis for translation and interoperability, not just an assumption that everyone’s models will happen to agree. It also normalizes impact across many levels at once — individual skills, relationships and trust, behavior, systems change, community and population-level outcomes — the holistic picture funders increasingly ask for, and the one communities themselves often already sense but can’t yet show.

Think of this as a concrete proposal for the “intelligence layer” gap described in connecting the tools, and a cousin of the questions raised in AI for participation.

Sources

• Civic Intelligence Infrastructure — Concept Overview (Brandon Nørgaard, Mediators Foundation, 2026).
• Brandon Nørgaard — OpenCivics Civic Innovator Session (2026): youtu.be/Umpu1AmTuy8.