Why campaign scale is where agents earn their place.
An issue campaign aimed at Congress has 535 voting members to account for: 435 in the House and 100 in the Senate. Each office has its own record on the issue, its own staff, its own committee seats and its own constituents among your members. Add a coalition of partner organizations, each with relationships of its own, and a campaign that runs for months, and the same piece of work repeats hundreds of times: read this office’s record, decide whether it is a priority, prepare the advocate who will meet it, track what happened.
That repetition is what AI agents suit. An agent is software that works through a multi-step task with some independence: it reads sources, compares them, drafts, updates a record and proposes what to do next, then hands the result to a person. For one office, a staffer can do all of that in an afternoon. For 535 offices, every week, nobody can, so most campaigns work the few dozen offices they know and guess about the rest.
Agentic advocacy lets agents carry that breadth while people keep every decision about who is asked, what is said and who says it. The board below runs one sample campaign through four agent passes. Watch where it stops.
535 offices, four agent passes, one human approval.
A board of all 535 congressional offices, 435 in the House and 100 in the Senate, for one sample issue campaign. Agent passes sweep across the board: research every office, prepare briefs for priority offices, monitor for new statements, and propose next actions. Proposed actions wait for a person. Only a person’s approval moves an office to engaged.
ReadyOne sample issue, the Rural Telehealth Access Act, across every office in Congress. Run the agent passes, or let them play.
- Status
- Not started
- Record
- Not researched yet
- Committee
- Not on the committee
- Advocates
- No advocates on file
- Coalition
- Summit Nurses Council
- Next
- Not researched yet.
Tap any mark to open its office.
What AI agents do across an advocacy campaign.
Each pass is the same job done for every office, with a person checking the work at the point where it matters. The useful question for any agentic advocacy tool is not what the agent can do, but what the person sees and decides at each step.
PASS 1Research every office
THE AGENTReads each office’s public record on the issue: votes on related bills and amendments, cosponsorships, floor and hearing statements, letters and committee seats. It notes the staffer who covers the issue where that is public, and attaches a source to every finding.
THE PERSONDefines the issue precisely (bill numbers, related amendments, the terms that matter) and spot-checks a sample of offices against the sources before anyone relies on the board.
PASS 2Brief the advocates
THE AGENTPrepares a brief for each priority office, written for the advocate who will be in the room: the member’s record in plain words, the ask, how the issue lands in that district or state, and what to avoid saying.
THE PERSONApproves the brief format and the ask, reads the briefs for the offices that matter most, and decides which advocates are matched to which meetings.
ALONGSIDETrack the coalition
THE AGENTKeeps the sign-on list current, records which partner covers which office, compares partner positions provision by provision, and flags duplicate or conflicting asks before they reach an office.
THE PERSONThe coalition lead decides who owns each office and settles conflicts. No organization is listed as a signer until it has said yes.
PASS 3Monitor statements and votes
THE AGENTWatches for new statements, cosponsor changes, votes, hearing remarks and staff departures, matches each to its office, and updates the board and every brief that depends on it.
THE PERSONReviews anything that changes the plan: a key member moving, a new opponent, a vote scheduled earlier than expected.
PASS 4Propose the next best actions
THE AGENTProposes the next step for each office, such as an advocate from home requesting a meeting, a thank-you after a cosponsorship, a partner follow-up, or waiting. Each proposal carries its reason and its draft.
THE PERSONApproves, edits or rejects every proposal. Staff send what is approved. The agent sends nothing.
One coordinated ask per office.
Coalitions are where scale bites hardest. Six partners each working from their own relationships will cover the offices they know best, which tend to be the same offices. Some offices hear from three partners in one week with three slightly different asks; others hear from nobody.
An agent that can see the whole coalition’s coverage, with the partners’ agreement, can propose a split: one owner per office, chosen from whoever already has the relationship, with the uncovered offices shared out. It can also hold the sign-on list and compare positions provision by provision. The coalition lead makes the call.
Six sample coalition partners and forty committee offices. When each partner picks its own offices, some offices are asked by several partners and others by none. When an agent proposes one owner per office, based on existing relationships, every office gets exactly one coordinated ask, and the coalition lead approves the split.
Each partner picks its own offices12 offices are asked by two or more partners, often for slightly different things, and 12 get no ask at all. Rings around an office show extra asks; hollow marks are offices nobody covers.
- MHA Meridian Health Alliance
- LPN Lakeshore Patient Network
- PRC Plains Rural Clinics Coalition
- HCA Harbor Caregivers Association
- SNC Summit Nurses Council
- NFP Northfield Family Physicians
The lines agents do not cross.
These hold at any scale. They are what keeps an agentic campaign an advocacy campaign, rather than automated contact that offices learn to ignore.
Every message to an office, an advocate or a partner is approved by a person as the exact text that will be sent. Approval is per message, not a standing permission.
An agent never writes as an advocate, signs as one or invents one. Constituent contact comes from constituents, in their own words.
No emails, no congressional web forms, no social posts, no calls. Agents prepare; people send.
Coalition members share what they agree to share. An agent working for one organization does not see another’s member list.
A person decides how often an office or an advocate can be asked. Agents flag any plan that would go past it.
A brief that tells an advocate how a member voted links to the vote. If the agent cannot find a source, it says so rather than guessing.
What goes wrong when agents run at campaign scale.
A mistake made once is an awkward meeting. The same mistake made across 535 offices is a pattern, and offices remember patterns. These are the ones to test for.
Hill staff move often. A brief that names a staffer who left last month tells the office nobody is paying attention. Check staff at every pass.
Matching advocates to offices by address is easy to get slightly wrong, and then wrong everywhere. Spot-check matches before a fly-in.
Tailored briefs slide back toward one template. Read a handful side by side; if they read the same, they are not tailored.
Agents make another touch cheap to propose. Offices and advocates notice frequency long before content. Set limits and keep them.
Without a view of the whole coalition, partners repeat each other. Check coverage before a week of meetings, not after.
An agent states a wrong vote as calmly as a right one. Keep sources attached so errors can be found, and review a sample at each pass.
Agents inside the back office, approvals with your people.
Capiro is the AI back office for government affairs. Meri, the AI Chief of Staff inside the back office, carries this kind of work across many offices and partners: reading congressional activity against your issue, keeping every office, staffer and coalition partner on one relationship map, drafting briefs with sources, and proposing next steps with their reasons.
Outreach goes out only after a person approves the exact message, and nothing is sent in anyone else’s name. Your advocate action tool still carries advocates’ own messages. AI should not replace the relationship. It should carry the work around it.
Frequently asked questions
What is agentic advocacy?
The use of AI agents, software that works through multi-step tasks with some independence, to carry the repetitive work of an issue campaign across many offices and partners: researching records, preparing briefs, tracking coalition sign-ons, monitoring statements and votes, and proposing next steps. People keep the decisions and all outreach.
How do agents keep coalition partners from asking the same office twice?
They work from one shared view of the campaign. Before outreach is queued, the agent checks which partner already owns each office and proposes one owner per office, plus the offices nobody covers. The coalition lead approves the split, and every message still goes out only after a person approves it.
How do agents handle coalition partners’ data?
Only with each partner’s agreement. A coalition can share office coverage, positions and sign-on status without sharing member lists. An agent working for one organization should see only what the other partners have chosen to share.
What should a person check before an agent-prepared brief reaches an advocate?
That the office’s record is right and sourced, that any staffer named still works there, that the ask matches the current position, and that the local angle is accurate for that district or state. Check a sample at every pass, and every brief for the offices that matter most.
Do AI agents replace grassroots advocates?
No. Agents make it possible to prepare every advocate for every office. The persuasion still comes from constituents speaking for themselves, and an agent must never pose as one or invent one.
Agents for the breadth. People for every ask.
Capiro is the AI back office for government affairs. Meri carries the research, briefing, tracking and monitoring across every office and partner in a campaign, and every outreach waits for a person to approve it.
The category, explained: What is an AI back office for lobbyists?
