Paperclip for Politics: AI Agent Orchestration for Political Campaigns

2. 1. The Problem: Campaigns Are Always Under-Resourced

Every political campaign, at every level, operates with fewer people than it needs. A Senate campaign with $5M raised still has maybe 30 full-time staff trying to cover an entire state. A state legislature campaign might have one paid staffer and a rotating cast of volunteers who need constant re-onboarding.

The result is that campaigns are always triaging. The press release doesn't go out because the comms person is doing voter file pulls. The research brief never gets written because the research director is fielding media calls. The fundraising email is late because everyone is focused on the field program.

This is structural, not a talent problem. Campaigns have hard budget constraints, a fixed election date, and a sprawling set of tasks that don't shrink. Something always gets dropped.

AI agents can run the stuff that gets dropped. Not the strategy, not the judgment calls, not the candidate prep. But the production work: drafting, researching, scheduling, monitoring, reporting. The bulk of what a campaign staff actually does with its hours.

3. 2. The Concept: What This Product Actually Is

The product is a Paperclip-style multi-agent orchestration platform, rebuilt around the mental model of a political campaign instead of a business.

In Paperclip, you define an org chart (CEO agent, COO agent, marketing agent), give each agent a mission and budget, connect them to tools (Claude Code, Cursor, HTTP endpoints), and let them run on a heartbeat schedule. The platform handles task queues, prevents duplicate work, logs everything, and gates anything sensitive behind human approval.

The political version replaces "CEO/COO/marketing" with "campaign manager/comms director/ field director/research director/digital director". The agents' tools are political: voter file APIs, press databases, FEC filing systems, social media APIs, canvassing platforms. The budget controls now also enforce FEC compliance. The approval gates are now also a review step before anything goes public under the candidate's name.

Think of it as giving a 3-person campaign the operational output of a 15-person campaign, while keeping humans in control of all judgment calls.

4. 3. Target Buyers

Primary: Mid-tier campaigns (state legislature to U.S. House)

These campaigns have real budgets ($200K to $3M), real stakes, and real staffing gaps. They're too small for an enterprise political consulting firm but too big to run on vibes. A House campaign with 8 staffers trying to compete with an incumbent with 20 is the exact buyer. They will pay $2K-$5K/month without blinking if it actually closes the gap.

Secondary: Advocacy organizations and PACs

A labor union's political arm, an environmental advocacy org, a single-issue PAC. These organizations run year-round, not just in election cycles, which solves the seasonality problem (more on that below). They have the same operational structure as a campaign (comms, research, field, digital, fundraising) and the same chronic understaffing relative to ambition.

Tertiary: Downballot and hyperlocal campaigns

City council, school board, county commissioner. Almost no paid staff, massive lists of tasks. These campaigns are currently underserved by political tech entirely because the price points are too high and the tools too complex. A $99/month tier with 2-3 agents could unlock a market that existing vendors ignore.

Not a fit: Presidential and statewide mega-campaigns

A $200M presidential campaign has 500 staff, bespoke tech infrastructure, and dedicated AI teams already. You're not selling to them. They'll build their own. That's fine.

5. 4. The Existing Political Tech Stack

Political tech is surprisingly fragmented. Campaigns stitch together 6-10 tools that don't talk to each other, and most of them are decade-old software that looks it.

No tool in this stack uses AI orchestration. Most barely use AI at all. The existing vendors are building incremental AI features into their tools (NGP VAN adding AI-generated call scripts, etc.) but nobody is building the layer that coordinates across all of them.

The product doesn't need to replace any of these tools. It sits on top of them, orchestrating agents that use them as inputs and outputs. The campaign keeps NGP VAN. The research agent just pulls from it.

6. 5. The 5 Core Agent Roles

The org chart ships with 5 default agent roles. Campaigns can add, remove, or customize.

Comms Director Agent

Drafts press releases, talking points, rapid response memos, and op-eds. Monitors press coverage and flags mentions of the candidate and opponent. Drafts responses to media inquiries for human review. Tracks the opponent's public statements and surfaces conflicts with prior statements. Output: everything goes to a human approval queue before it's ever sent.

Research Director Agent

Tracks legislative votes, public records, FEC filings, court records, and social media history for both the candidate and the opponent. Builds and maintains a living research brief. Surfaces new developments daily. Flags when the opponent says something inconsistent with their record. The most valuable agent for campaigns that can't afford a full-time researcher.

Field Director Agent

Generates canvassing scripts tailored to voter segments and issues. Drafts volunteer recruitment emails. Creates training materials for new volunteers. Analyzes field report data from NGP VAN and surfaces trends (which neighborhoods are underperforming, which scripts are getting good responses). Does not replace the field director's judgment on strategy, just handles production.

Finance Director Agent

Drafts fundraising emails, texts, and social appeals. Tracks fundraising progress against goals and drafts deadline urgency copy. Monitors major donor giving history and drafts personalized outreach for human review. Flags potential FEC compliance issues before they become filings. Does not have the ability to send anything. All outreach is human-approved.

Digital Director Agent

Drafts social media content for all platforms. Schedules posts via integrations. Monitors social engagement and surfaces trending content relevant to the campaign. Drafts ad copy variations for testing. Analyzes what's performing and recommends pivots. Everything goes to a human review queue before posting.

7. 6. Key Product Features

Campaign org chart

Visual org chart with hierarchies, roles, and reporting structures. Campaign manager (human) at the top. Agent directors below. The campaign manager sets the mission, priorities, and constraints. Agents execute within them.

Heartbeat scheduling

Each agent runs on a configurable schedule. Research agent checks news and filings every morning. Digital agent drafts social content every evening. Finance agent drafts the weekly fundraising email every Thursday. No human needs to kick off routine work.

Approval queue

Everything a campaign publishes under the candidate's name goes through a human approval queue. This is non-negotiable. The product should never have a pathway for automated public output without human sign-off. Not just a UX choice, a legal one.

Budget controls

Per-agent monthly token budgets. Total campaign budget cap. Real-time spend visibility. Campaigns run on tight budgets, they need to know exactly what AI is costing them.

FEC compliance layer

A specialized module that flags potential compliance issues. Contribution tracking rules, disclosure requirements, coordinated expenditure limits. Not a legal service, a first-pass filter. "This looks like it might be a problem, talk to your compliance counsel."

Full audit log

Every agent action, every draft, every approval, every rejection, logged immutably. Campaigns sometimes get subpoenas. They need records.

Template library

Pre-built campaign templates by race type (House, state senate, city council, school board) and by party. Import a template, customize the candidate details, and the agents know what they're working toward. Lowers the setup time from days to hours.

8. 7. Monetization

Subscription tiers

Seasonality and the advocacy org answer

The obvious problem: campaigns are seasonal. A U.S. House campaign pays for 18 months, then it's over. MRR spikes in election cycles and craters off-years.

The answer is advocacy organizations, PACs, party committees, and think tanks. These operate year-round with identical operational structures. Selling to 200 advocacy orgs at $1,999/month is $4.8M ARR with no seasonality. Campaigns become the high-volume, high-churn top-of-funnel that eventually upgrade some customers to permanent org accounts.

Unit economics

Political campaigns have real budgets and known willingness to pay for operational tools. A $2M House campaign spends $200K-$400K on staff. Replacing even 10% of that output with AI agents at $24K/year is a no-brainer sale. The ROI conversation is easy. The compliance and trust conversation is harder, but solvable.

9. 8. Risks and Hard Problems

The misinformation liability

The hardest problem. If an AI agent drafts a press release that contains a factual error and a campaign publishes it without catching it, who is liable? The platform? The campaign? This is both a legal and reputational risk.

Mitigation: mandatory human approval for all public output, clear ToS that the platform is a drafting tool not a publishing platform, required sourcing links in all research output so humans can verify before approving.

Astroturfing and abuse

The platform could be used to generate coordinated inauthentic content at scale. Fake grassroots. Sockpuppet armies. This is a real risk.

Mitigation: require verified campaign registration (FEC filing number or state equivalent) to access the platform. Rate limits on content generation. Terms of service that explicitly prohibit coordinated inauthentic behavior. Cooperate with election authorities if abuse is found. Not a perfect solution, but establishes legal and moral defensibility.

FEC and election law complexity

Election law varies by state, changes frequently, and has real criminal penalties. Building any compliance layer is a liability if campaigns rely on it and it's wrong.

Mitigation: be very explicit that the compliance module is not legal advice. Recommend campaigns have human compliance counsel. Flag issues, don't resolve them.

Seasonality

Already addressed: anchor on advocacy orgs for stable ARR, use campaigns as high-value burst revenue.

Trust and the "AI wrote that" attack

In politics, an opponent discovering that a campaign uses AI to draft communications becomes an attack ad: "Candidate X uses robots to talk to voters." This is a real political risk for buyers, not just a product risk.

Mitigation: position as "AI-assisted, human-approved" not "AI-generated." Every public output goes through a human. The human is authoring the final product. The AI is the research assistant and first draft, like every journalist uses today.

10. 9. Go-to-Market Path

Phase 1: Advocacy orgs (year 1)

Don't start with campaigns. The sales cycle is too short (campaigns end), the trust is harder to build fast, and the seasonality kills MRR stability. Start with progressive or conservative advocacy organizations that run year-round and have clear, recurring operational needs. Target organizations with 5-50 staff where an AI operations layer is clearly additive.

5-10 design partner organizations, deeply involved in product development. Build around their actual workflows. Get them to $99-$499/month first, expand to $1,999+ once the value is obvious.

Phase 2: Downballot campaigns as top-of-funnel (year 1-2)

City council and state legislature candidates are ignored by existing political tech vendors. The tools are too expensive and too complex. A $99/month product designed for a 3-person campaign team is a genuinely underserved market. These campaigns are high-volume (tens of thousands run every cycle), they talk to each other, and they graduate to bigger races. Build community and brand at this level.

Phase 3: Mid-tier campaigns (2026 election cycle)

With a track record from advocacy orgs and downballot campaigns, approach U.S. House campaigns and competitive state senate races in the 2026 cycle. These campaigns have budgets, real pain, and clear benchmarks for ROI. The $1,999-$4,999/month tier becomes the core revenue driver.

Distribution

Political tech spreads through consultants. Campaign consultants recommend tools to every campaign they work with. One good relationship with an active political consultant can seed dozens of campaigns. Build a partner/referral program for political consultants from day one.

Also: political training programs (campaign schools, party training programs) are an underrated distribution channel. If you're the tool they learn in training, you're the tool they use when they run a real campaign.

11. 10. Verdict

This is a real problem with a clear product shape. The pain is acute and structural, the existing tooling is visibly weak on AI orchestration, the target buyers have known budgets, and the Paperclip metaphor maps cleanly onto campaign org structure.

The hard parts are real too. Misinformation liability, abuse potential, and election law complexity are not hand-wave-away problems. They require thoughtful product decisions (mandatory human approval, verified registrations, explicit compliance disclaimers) that need to be baked in from day one, not bolted on later.

The biggest strategic question: do you build this as a Paperclip fork (open-source core, political tooling on top) or as a purpose-built product? The Paperclip fork approach is faster but technically messier. Purpose-built is slower but lets you design around campaign workflows from scratch.

Given the compliance and trust requirements of political tech, purpose-built is probably the right call. You want every design decision made with "this is for a political campaign" as the first constraint, not retrofitted in.

The advocacy org anchor for stable ARR is the key insight that makes this a real business and not just an election-cycle play. Build that base first, then ride campaigns as a high-value growth lever every two years.