product
OpenCouncil serves three audiences, in priority order:
- Citizens — residents of Greek municipalities who want to know what their council decided and why. They arrive from a search result, a shared link, or a notification; they are not trained users and may visit once a month or once a year. Many are older, on mobile, and reading Greek.
- Journalists & civic watchdogs — power users who search across meetings, follow specific people, parties, and topics, and need to cite the record precisely (timestamps, word-for-word transcripts, links to official Diavgeia publications).
- Municipal staff & council members — administrators who manage meetings, review transcripts, correct speaker attribution, and release content to the public.
The job to be done on any given screen: find what was said, who said it, and what it means — with the least possible friction between a citizen and the public record.
OpenCouncil digitizes, transcribes, and makes municipal council meetings searchable, helping citizens engage with their local government. It is built by Schema Labs, a non-profit building technology to strengthen democracy. The platform turns hours of council video into word-for-word transcripts with speaker recognition, AI summaries, subject categorization, full-text search, personalized notifications, and shareable highlights — all open-source, with open data via a public API.
Success looks like: a citizen finds the exact moment their street was discussed in under a minute; a journalist quotes the record with confidence; a municipality trusts the platform enough to make it their official record of proceedings.
Civic, trustworthy, approachable.
A serious public-interest tool that ordinary citizens can use without training — credible but warm. The voice is plain-spoken Greek (and English), never bureaucratic, never breathless. The interface should feel like a well-run public institution staffed by people who actually want you there: rigorous about the record, generous in how it explains it. AI-generated content (summaries, categorization) is always clearly labeled and subordinate to the verbatim record.
Visual patterns OpenCouncil must avoid:
- Bureaucratic portal density — dense nested menus, PDF-first layouts, institutional gray-blue palettes, and walls of unbroken administrative text. These patterns make public records hard to access; OpenCouncil exists to make them legible.
- Marketing-site gloss — decorative gradients, oversized hero metrics, repeated identical card grids, and conversion-funnel landing patterns. OpenCouncil is civic infrastructure, not a growth product.
- Visual urgency and clutter — stacked banners, multiple competing accent colors, and persistent urgency cues. OpenCouncil communicates through accuracy and clarity, not visual volume.
- The record is the interface. Transcripts, votes, and decisions are the content; the UI recedes so the public record can carry the weight. Density is acceptable; decoration is not.
- Trust through fidelity. Word-for-word comes first; AI assistance (summaries, subject extraction) is clearly labeled, never silently blended into the verbatim record. Timestamps, sources, and Diavgeia links are first-class.
- Every citizen, not just the engaged ones. Screens must work for a 70-year-old on a phone as well as a journalist on a desktop. Plain language over civic jargon; one obvious action per screen.
- Earned familiarity. Use standard, well-understood affordances (search, tabs, lists, players). Novelty spends trust this product cannot afford.
- Warmth in moments, rigor everywhere. The orange identity shows up at decisive moments (primary actions, highlights, the brand) — not spread across every surface.
- WCAG 2.1 AA is the commitment: ≥4.5:1 body-text contrast, full keyboard navigation, visible focus states, reduced-motion alternatives for every animation.
- Multilingual by architecture (next-intl); Greek is the primary locale, English second, with more languages planned for multicultural cities.
- Audiences skew broad in age and technical ability — touch targets, font sizes, and reading level should assume the least-technical citizen, not the median SaaS user.