For the last year we’ve been building the thing we wished existed when we were running races with a CSV, a Google Sheet, and a prayer. Today we’re opening up Civitas to a wider beta.
Here’s what it is, how it fits into the broader ConstituencyData project, and how to get access.
The short version
Civitas is an AI-native workspace on top of the ConstituencyData catalog. You open it in a browser, you ask questions about your district in plain English, and it builds the chart, pulls the list, scores the donors, routes the walk, or stress-tests the mailer — right there, without a BI analyst in the middle.
Four surfaces, one workspace:
- Ask Civitas — chat interface over voter, donor, and government data.
- Donor Board — call-time ranked list with propensity × expected-gift scoring.
- Walk Planner — field ops routing with home + business dual-plot and business-hour awareness.
- Message Lab — paste any asset (mailer, email, ad, press release), get a stress test back.
It’s one tool. It behaves like one tool. You are not tab-hopping between four vendors.
Why AI-native
A quick note on what we mean by “AI-native,” because the phrase has been cheapened.
An AI-native tool is one where the primary interface is natural-language conversation with your data, and the underlying system is honest about what it knows and doesn’t. That honesty matters. A lot of “AI features” bolted onto legacy tools are wrappers that occasionally hallucinate a voter count and ship the result to the user. That’s worse than no feature at all.
Civitas is built so that every number, list, or chart you get out is grounded in an actual SQL query against an actual table. The query is visible to you. If the model doesn’t know, it says so instead of guessing. Generative outputs (draft a mailer, summarize a region) are clearly marked as generative, and grounded queries are clearly marked as data. You always know which is which.
We route each query to the best model for the job. Long-context and summarization lean on Gemini. Nuanced reasoning and writing lean on Claude. Statistical reasoning goes to reasoning-tuned models. On the Max plan you get frontier tier by default. This means Civitas benefits from every frontier advance in the lab without you having to pick a model.
Who it’s built for
We’ve said this in a few places, but it bears repeating. Civitas is built explicitly for:
- State and local campaigns — school board, city council, state house, county commission, state senate. Races where the legacy vendor price point is bigger than the campaign budget.
- Consultants running multiple races — white-label outputs, multiple workspaces, client sharing.
- PACs and advocacy organizations — issue-universe building, coalition mapping, donor research.
- Marketing and PR teams with political adjacency — regional targeting, message testing, sentiment research.
We are deliberately not building this for presidential campaigns. They have vendors. Those vendors are fine. We’re not trying to displace them; we’re trying to extend the infrastructure to the 95% of races below that level.
What’s in the beta today
Functional and tested in the beta:
- Chat over the full ConstituencyData catalog (voter file, FEC, state donors, business licenses, government records, election results)
- Chart builder (bar, line, scatter, choropleth) from natural language
- Saveable query library per workspace
- Donor propensity and expected-gift scoring (NV live, FEC federal live)
- Walk planner with business-hour routing
- Message stress test (mailer, email, social, ad, press release)
- CSV / XLSX / PDF / PNG exports within plan limits
- Workspace sharing for multi-seat accounts
Rolling out over the next 90 days:
- Yard-sign placement scoring
- White-label PDF briefings (Max tier)
- Civitas API (Max tier)
- Meta tribe-model integration (when Meta ships)
- Additional state donor files (NY and CA next in priority order)
Founder pricing
Beta users lock in founder pricing on Pro and Max for two years. Pro is $99/mo at launch; founders stay at $99/mo through 2027 even if pricing moves. Max is $299/mo at launch, same deal.
The Free tier is free forever, beta or not.
Getting in
You can sign up at any time — there’s a form at the bottom of the Civitas page. We onboard in waves, primarily to keep support quality high during the early weeks. Most signups get their workspace provisioned within seven days.
If you have a time-sensitive race (primary in Q2, special election on the books), tell us when you sign up and we’ll prioritize.
What we need from early users
The beta is genuinely a beta. We have strong unit testing, a solid model evaluation harness, and good observability — but there’s no substitute for a hundred campaigns banging on the thing in anger. What we specifically need from early users:
- Queries that confused the model. If you asked something reasonable and got a weird answer, tell us. We re-train on real examples.
- Edge cases in the data. If you find a record that looks wrong, a duplicate we didn’t catch, a geocode that’s off by a block, tell us.
- Feature gaps. If there’s a question you’d ask your expensive vendor’s account rep, we want to know what that question is.
- Legal or compliance concerns. If a query feels like it should be rejected and wasn’t, or if something was rejected that should have been permitted, tell us.
Send everything to hello@constituencydata.com. We read all of it.
Closing thought
This product only exists because the infrastructure for building it — frontier LLMs, cheap geocoding, open data, modern web platforms — matured fast enough that a small team could credibly compete with a twenty-year-old industry. We don’t take that for granted. We intend to earn your trust by being useful, being transparent, and keeping the pricing where a real campaign can actually reach it.
Welcome to Civitas. Go run a good race.