AI Made Production Cheap. Mid-Decade Redistricting Just Made Data Expensive.

Between July 2025 and yesterday afternoon, seven states finalized or announced mid-decade congressional maps — Texas, Missouri, North Carolina, Ohio, Utah, California, and now Florida. Texas alone created roughly five new Republican-leaning districts. Missouri redrew Kansas City into surrounding rural districts to net a single seat. The Cook Political Report's national read across all of it is roughly a wash. The political coverage is fixated on the seat-count math, and the math matters. The campaign-operations math is different and is barely being written about. Every campaign running in a redrawn district just inherited a partial-strangers problem at the data layer — and that problem is what is going to decide a meaningful share of the contested House races in November.

Production Is the Wrong Constraint to Be Optimizing For

The other story running in the political press this month is the AI story. The NRSC is using AI to create ads, analyze data, and identify new small donors. A half-dozen AI-generated spots are already running. Axios reported a 64–49 split in daily AI use among Republican and Democratic consultants, and a March video took a single print quote into 1.7 million views at effectively zero cost.

The operational read is simple. The marginal cost of producing a campaign message — a digital ad, a fundraising email, an SMS variant, a piece of creative tailored to a microsegment — is now near zero. Production was the constraint for thirty years of campaigns. It is not the constraint anymore. When production stops being the constraint, the next constraint is the data the production is targeted against — the list, the segmentation, the deliverability infrastructure. Those are the layers mid-decade redistricting just put under stress.

Mid-Decade Redraws Break Three Things at the Data Layer. None of Them Are the Creative.

The first is the voter file. Every voter in a redrawn district has to be reassigned to the new boundary, and the rebuild takes weeks. Precinct boundaries change with every cycle, geocoding fails cleanly on a small but real share of residential addresses, and what comes back is a freshly assigned universe with imperfections at the edges. Clean targeting goes dark for the duration.

The second is the candidate's owned list. The email list, SMS subscribers, donor file, and volunteer roster were all built around the old district. After the redraw, a meaningful share of those records are people who no longer live in the district. They may still donate. They may still volunteer. They are no longer voters in the candidate's race. Most campaigns will not segment cleanly against the new geography and will keep mailing the entire list — accelerating list fatigue and depressing performance with the in-district contacts who actually matter.

The third is issue-affinity segmentation. The issues that ranked in a 2024 district map don't necessarily rank in the 2026 version. A district that was 60 percent suburban and is now 40 percent suburban needs a different issue mix than the one the campaign refined over the last two cycles. Targeting attributes only earn their lift when applied to a district whose composition has been re-baselined.

Production Is Cheap. Trust Is Still Earned.

The asymmetry is what gets missed. AI can produce two hundred variants of a fundraising email tailored to two hundred microsegments before lunch. It cannot produce engagement that didn't happen. The opt-in supporter who has signed a petition, opened the campaign's email, replied to an SMS, donated twenty-five dollars, or attended a virtual event has given the campaign a relationship the model did not have to invent. Everyone else is a guess.

For any campaign in a newly drawn district, the share of the new voter universe with which the campaign has observed engagement is small. AI cannot fix that, because AI cannot generate behavior that did not occur. It can only generate creative for engagement the campaign still has to earn. That is the work the next twelve months are for.

The Twelve Months Ahead Are What Matter Now

The map is the starting line. The campaigns that get the most out of the redraw will treat the next twelve months as a foundation period and approach the new district deliberately, before production volume accelerates further.

Start with the owned list. Audit it against the new boundary, separate the in-district contacts from the out-of-district ones, and treat them differently from the next send forward. Out-of-district donors and volunteers are still real assets — they are not voters in the new race, and continuing to mail them general-election turnout content burns the relationship and the sender domain.

Re-baseline issue affinity from the ground up. The composition of the new district is the new ground truth. The issue mix that performed in the prior version is a starting hypothesis, not a conclusion. New surveys, fresh opt-in opportunities, and tagging based on observed behavior are how the segmentation gets rebuilt to match the geography.

Then earn engagement on every new contact, one record at a time. Clean opt-in capture, warming sequences that earn the right to ask before they ask, and a sender reputation protected by what the campaign chooses not to send. We have built programs that took an engaged audience from about two thousand contacts to more than fifty thousand in twelve months — not by sending more, but by being ruthless about relevance. The mechanism is straightforward and it travels: relevance compounds, and a list built on observed behavior holds up when the geography around it changes.

AI made the production cheap. The redraw made the data expensive. The work of the next twelve months is the work that decides what comes back when the production hits send.