Executive overview
Duly is a civic engagement platform that verifies constituent identity, turns natural-language input into structured insight for staff, and publishes anonymized district trends so the public can see where issues stand.
- Verified voices: residency check + one-time code; district routing.
- AI summaries, not decisions: topic, sentiment, urgency, deduplication; staff stay in control.
- Transparency: a public portal shows anonymized issue trends and official responses.
- Privacy by design: no public PII; redaction and audit trail across the pipeline.
Problem
Offices are swamped with unverified, unstructured messages. Real constituent voices are diluted; staff rely on filters and scripts; the public can’t see if concerns led to action.
- Fragmented channels: email, web forms, calls, social DMs — no unified intake.
- No verification: offices can’t trust residency or constituency at scale.
- Low signal-to-noise: mass campaigns/bots bury everyday residents.
- Opaque outcomes: citizens rarely see if/when an issue is addressed.
Who it serves
Constituents
- One verified place to be heard by the correct office.
- See how your input contributes to district trends.
- Opt-in follow-ups when leaders respond or act.
Politicians & staff
- Noise-free inbox: verified messages, clustered by topic/sentiment/urgency.
- Dashboards to detect emerging issues and track response SLAs.
- Publish official responses to the public portal with one click.
General public
- District-level, anonymized summaries of what people raise.
- See responsiveness over time without exposing personal data.
Product principles
- Verified voices only: input counts after residency verification.
- Privacy-first: PII is redacted; public data is aggregated and anonymized.
- Transparency without partisanship: issue- and data-driven, not editorial.
- AI as a tool: assists with summaries; humans decide and publish.
- Build for trust, not engagement metrics: fast, respectful UX — finish in under 5 minutes.
What ships in the MVP
Included
- Constituent verification (ZIP + email/SMS; optional higher assurance).
- Message intake (structured prompts, free text, voice-to-text).
- AI summarization: topic, sentiment, urgency; deduplication and spam guard.
- Staff dashboard: clusters, filters, message review, “needs reply/escalate”.
- Public portal: anonymized trends, published responses.
- Notifications: follow-ups on topics users care about.
- Accessibility (WCAG 2.1+), mobile-first layout.
Out of scope (Phase 2+)
- Mass outreach campaigns, real-time chat.
- Advanced analytics beyond core trends.
- Full voter-roll integrations; multilingual beyond English.
How it works (pipeline)
Verify the sender → structure the message → route by policy/SLA → staff act → publish anonymized outcomes.
For staff
- Verified, deduped inbox with safe-handling lane for threats.
- Clusters show volume, sentiment, subtopics, and urgency.
- Publish official responses to close the loop publicly.
For constituents
- Simple input; clear confirmation; optional follow-ups.
- See how your message contributes to district-level trends.
Success metrics (pilot targets)
↑ verification
% verified users submitting monthly
< 24h
Time-to-first-response
−50%
Time-to-resolution
30–60%
Automation/deflection on common requests
↑ transparency
% feedback summarized & published
Additional measures: NPS (constituents & staff), model accuracy vs. human review, public dashboard engagement.
Architecture & feasibility
High-level design (NDA-safe)
- Multi-tenant: strict tenant isolation; scoped access; per-district config.
- Verification: ZIP + email/SMS; optional higher-assurance integrations later.
- AI pipeline: topic clustering, sentiment/urgency, dedup/spam; human-in-the-loop for pilots.
- Governance: PII redaction, hashed IDs, immutable audit/export; WCAG 2.1+.
- Public data: anonymized aggregates streamed to the transparency portal.
Risks & mitigations
- Identity & integrity: start simple; explore voter-file partnerships; log metadata for audit.
- AI accuracy: human review in pilots; confidence scores; flagging to improve models.
- Adoption: begin with willing offices; deliver clear value (weekly trend briefs).
- Spam/abuse: verification, rate limits, heuristics; safe-handling lane for threats.
- Legal/ethical: privacy counsel; opt-in transparency; third-party audits where applicable.
What I owned
- Product thesis, principles, and MVP scope; stakeholder interviews.
- Architecture and data governance; verification strategy; AI prompting & evaluation.
- UX for intake, cluster review, public transparency, and audit logs.
- Delivery leadership: CI/CD, security posture, incident playbooks, pilot success cadence.
Pilot with Duly
We’re inviting 5–10 additional pilot municipalities. If verified voices and transparent follow-through would help your community, let’s talk.