november keeps what the candidate has said, listens to the race, drafts cited responses, puts every one in front of the people who decide, and hands volunteers something true to share. Here is each part, with the screen it happens on.
Every speech, town hall, interview and strategy meeting goes into one record, transcribed and split into what the candidate actually said. Each claim, commitment and position is pinned to the span of transcript it came from. When anyone on the team, or any agent, needs a fact, this record is the only place it comes from.
november follows what the opponent, the press and voters are saying, and gathers it into clusters by topic and by who is speaking. Each cluster is ranked by urgency and comes with a suggested action, so the team decides what to answer from a short, sorted list instead of a raw feed. If the AI is paused, keyword triage keeps the queue sorted.
From a cluster, november drafts a response in the candidate's voice in minutes, sentence by sentence. Every factual sentence cites the transcript it rests on: open a citation to read her exact words, highlighted, with when she said them. Guardrails check the draft as it is written, and anything the record does not support is left unsaid.
Nothing goes out because an agent decided it should. Every draft waits in a review queue, scored against the campaign's own KPIs (audience fit, brand, message, grounding, and how much of it AI wrote), and each score gives its reasons, sentence by sentence. Approval can require a second person, and every approval, return and override is written to an audit trail.
Approved material becomes a plain explainer page that a door-knocker can share with a QR code, and every statement on it opens its source. Voters can ask a question: november answers only from the candidate's record, with the sources. When the record does not cover it, the page says “I don't know” and passes the question to the campaign.
Released content goes to the campaign's amplifiers channel by channel, every variant taken from the approved text. Each volunteer gets a kit on their phone, ready to copy and share in one tap, and a board follows every assignment from kit sent to posted or skipped.
Reach and engagement are reported by publication and by amplifier, and every figure says where it came from: counted by november, entered by hand, or tracked through a volunteer's link. The demo's simulated figures are labelled simulated and can be hidden.
Each campaign stays sovereign, with its own record, its own people and its own approvals. A campaign can offer approved material to a peer it trusts, and the peer receives a frozen copy that publishes only after its own approvals. The network sees lineage, who shared what with whom, not the content.
In the demo, every screen names the job it was built for. Open the story overlay to read the situation that screen serves and its acceptance criteria, so you can check the tool against what it was asked to do.
The demo is a working console for a fictional race. Sign in and follow one response from the candidate's record to a published, cited answer.