An advocacy group's biggest enemy is not opposition, it is supporter drop-off between caring and doing. A visitor who reads a moving campaign page and feels ready to act needs a small, specific next step: call this senator, sign this letter, attend this town hall, donate to this fund. The default WordPress experience hides those next steps behind sidebar widgets and footer links. SleekAI reads campaign pages, policy briefs, and action toolkits from WordPress and answers "how do I help" with a concrete instruction tied to the visitor's location, language, or area of interest.

Position lookups are the other recurring task. Journalists, allies, and skeptical visitors want to know exactly what the group's position is on a specific bill, executive action, or court decision. The campaign team usually has a position written up, but it lives inside a brief PDF or a press release archive that nobody reads. Upload those to OpenAI Files and the bot can answer "what's your position on House Bill 4421" with the actual statement, cited to the press release, with a link to the full document. That is more useful to a journalist on deadline than any contact form.

Display conditions and multibot also matter because advocacy work tends to be episodic. A campaign cycle has a public-facing intake bot, an organizer-only bot scoped to logged-in volunteer captains, and a press bot scoped to journalists with a specific URL token. Each speaks differently, with different context, and each can be enabled or disabled per campaign without touching the others. Conversation logs sit in WordPress, so supporter and journalist questions never leak to a vendor.

Advocacy groups operate on a thin attention budget. A supporter who reads a campaign page and feels moved has maybe 90 seconds of momentum before life pulls them back to something else. The traditional response is a fat homepage with three calls to action competing for the eye, which usually means none of them land.

A chatbot that converts "how do I help" into a single specific next step, tied to the visitor's location or interest, wins back some of that momentum. The bot is not built to chat indefinitely; it is built to deliver one concrete action per conversation and then get out of the way. Position fidelity is the other half of the value.

Journalists, allies, and skeptical visitors all want to know exactly what the group's position is on a specific bill or court decision, and the answer usually exists in a press release or a policy brief that no one reads. Uploading those documents to OpenAI Files makes them queryable in plain language with citations, which is meaningfully more useful to a journalist on deadline than a press contact form. The press bot does not write new statements or speculate, it surfaces existing ones with the date and source attached.

That posture protects the group from misquotation while making its work more accessible. Operational separation is the third piece. Multibot and user-role display conditions let an organizer-only bot exist alongside the public bot without leaking.

A new volunteer captain can ask the organizer bot about phone-bank objection handling at 9 PM the night before a Saturday rally without bothering the volunteer coordinator. The public never sees that bot, the organizer never sees the public bot, and the press bot lives on its own quiet URL. Privacy posture matters here too because advocacy organizations have an obvious adversarial-surveillance threat model.

Running on your own API key, with logs inside your own WordPress database, is the right default for a group that may need to argue in front of a hostile jurisdiction someday.