Most winery leaders I work with have one of two content problems. Either they post too little because the blank page is exhausting and there is always something more urgent. Or they post enough but every piece sounds like it was written by a different person, because three different staff members each took a turn that week. Both problems are the same problem in different clothing. There is no Speechwriter on the team.
The Speechwriter agent is the second one I install on a new shadow team. It reads the Researcher's morning digest at 6:30 AM, checks the Connector's content calendar to see which audience today is for, and writes the day's draft against both. The draft lands in your review folder by 7:00. You open it with coffee, mark it up, and send it. Your weekly content rhythm now runs by itself.
This post is the per-agent breakdown for the Speechwriter in the Agent OS Webinar series. Week two of the 90-day install order. It depends on the Researcher's output, which is why it comes second.
What the Speechwriter reads (Inputs)
The Speechwriter has more inputs than any other agent in the chain. That is because writing in your voice is hard. Five categories feed it.
- Your voice bible. A single document — usually two to four pages — covering the rules your writing follows. Compliance phrasing you keep or avoid. Words that sound like you and words that do not. Sentence-length preferences. The greeting and signoff conventions for each channel. The voice bible is what makes the difference between a draft that needs five minutes of editing and one that needs a full rewrite.
- Your archive. The last 50 to 200 pieces of external content you have already published — blog posts, LinkedIn posts, customer-newsletter sends, press quotes, founder letters. The Speechwriter reads this archive to learn your cadence and to know what you have already said.
- The Researcher's morning digest. Today's three industry stories that matter for your operation. The Speechwriter uses the digest to decide what is worth writing about now.
- The Connector's content calendar. Today's audience. Tomorrow's audience. The campaign you are in the middle of. The Speechwriter picks the topic against an audience and an arc.
- Recent meeting notes and decisions. When you said something useful in a Tuesday meeting that should reach customers, the Speechwriter notices and proposes a draft. Your best material comes out of the conversations you are already having.
The voice bible is the input most CEOs try to skip. Skip it and the Speechwriter produces drafts in the AI-default voice that all six other businesses in your sector are also producing. Write the voice bible properly and the Speechwriter produces drafts in yours.
What it does with that material (Skills)
The Speechwriter has four skills, in rough order of usefulness.
- Voice match. Read the archive. Match the cadence, the sentence lengths, the channel-specific tone. A LinkedIn post is not a blog post. A wholesale-buyer email is not a tasting-room newsletter. The Speechwriter holds those channel-specific voices separately.
- Content history review. Catch repetition before it ships. The Speechwriter remembers that you wrote about your terroir three times this year already. It flags the new draft and proposes a different angle — usually a fresher one, drawn from this week's events rather than last quarter's reflections.
- News-driven drafting. Today's draft starts with one of the Researcher's three stories. The Speechwriter picks the angle that fits your operation, ties it to the audience the Connector flagged, and writes the piece. Speed matters; the news ages.
- Campaign and series planning. Once a month, the Speechwriter and the Connector draft the next month's content plan. A series on one theme over four weeks builds an audience. Twelve unrelated posts do not. The Speechwriter is the agent that holds the arc across a month.
The Speechwriter does not press send. That hard rail is shared with the Executive Assistant and the Personal CFO. External communication that goes out under your name needs your eyes on it before it ships.
The schedule (Scheduled Tasks)
The Speechwriter runs on a daily cadence plus a few weekly and monthly tasks.
- Daily at 6:30 AM. Reads the Researcher's overnight digest, checks the Connector's calendar for today's audience, drafts the day's content piece. The draft lives in your review folder by 7:00 so the Chief of Staff can include it in the morning briefing.
- On demand: pre-publish voice check. Paste a draft you wrote yourself, ask the Speechwriter to flag any sentence that does not match the voice bible. Three to five flags per page, usually with proposed rewrites. Useful for press quotes, board emails, anything where the voice has to be exactly right.
- Monthly drift sweep. Reads the last 30 days of published content and compares to the voice bible. Flags drift. If the voice has shifted, you find out before the customer does.
- Monthly content plan, paired with Connector. Next month's calendar and theme arc. One sitting, one document, ready before the first of the month.
- Quarterly campaign brief. A larger arc — a launch, or a multi-month series tied to a season. The Speechwriter drafts the brief; you edit it.
What lands on your desk (Outputs)
Three primary outputs, all stored in your vault so the next month builds on this month.
- The daily draft. One piece, in your voice, against today's audience, sitting in the review folder by 7:00 AM.
- The line-level flag list. For drafts you wrote yourself, a short list of sentences that read off-voice with proposed rewrites attached.
- The monthly content plan. Four weeks of topics tied to a theme arc, audience targeting from the Connector, and the Researcher's anticipated input categories pre-mapped.
The Speechwriter never sends external messages. It writes to your drafts folder and your content plan. The signing-off is yours.
The goal: every external word sounds like you
The Speechwriter exists for one outcome. Every external word — your LinkedIn posts, your tasting-room newsletter, your wholesale outreach, your founder letter — sounds like you, ships on a rhythm, and reflects what is happening in your industry today.
The compounding effect is sharp here. A Speechwriter that has been reading your archive for six months knows your sentence rhythms. It knows that you bury the lede sometimes and that this is intentional. It knows the words you do not use and the metaphors you reach for too often. The drafts get closer to your voice every month, because the archive of your published voice keeps growing.
The Speechwriter is also the agent that catches the slow voice drift no one else notices. Brands drift away from their voice the way a vineyard drifts away from its style — slowly, by a series of reasonable small decisions, each one fine on its own. By month nine, the company is writing in a voice nobody chose. A Speechwriter with the original voice bible loaded does not let that happen.
How the Speechwriter demonstrates Compound-the-Colleague
The Speechwriter is the canonical example of the Compound-the-Colleague teaching move. Two upstream agents feed it. Without the Researcher, the Speechwriter has nothing fresh to write about. Without the Connector, the Speechwriter has no audience or arc to write against. The drafts are useful because two other agents did their work first.
Here is what this looks like at Not Really Wines, our fictional demo winery at NunnCurtis Labs. NRW makes premium non-alcoholic wine, vacuum-distilled from Sonoma fruit, and ships DTC to forty-two states. Priya Sandhu, Director of DTC and The Society, drives the content rhythm. Her daily run looks something like this:
- 6:00 AM — The Researcher reads the wine industry sources, picks three stories that matter for NRW. One of them is about non-alcoholic wine sell-through in mass retail jumping sharply.
- 6:30 AM — The Speechwriter wakes up, reads the digest, checks the Connector's calendar. Today is a Society-subscriber day. It drafts a 200-word note tying the retail trend to NRW's own Society members — the people who have been on this train since before it was a trend.
- 7:00 AM — The draft lands in Priya's review folder.
- 7:30 AM — Priya opens her laptop with coffee, reads the draft, changes two lines, and queues it for the 11 AM newsletter send.
Priya did not start writing today's note. The Speechwriter did. The Researcher told it what was happening. The Connector told it who was reading. Priya did the editing pass that only a human can do — the one where you decide whether the draft is honest, on-brand, and worth your name. The full system is in place by 7:30 AM. The newsletter ships at 11. The rest of Priya's day is open for the work only she can do.
Why this is week two of the install order
The Speechwriter comes second for one specific reason: it depends on the Researcher. Build the Speechwriter first and it writes from a stale archive against a generic news landscape. The drafts feel close but not sharp. Build the Researcher first and the Speechwriter has fresh, filtered, operation-specific material to work with from day one.
Two more reasons:
- The Speechwriter ships your publishing rhythm immediately. In week one you have a digest. In week two you have a digest plus a draft a day. The pattern of content is built before you even reach the Chief of Staff or the Connector.
- The voice bible work is front-loaded but worth doing slowly. The Speechwriter is the agent that rewards patience on its setup. Spending three hours writing a careful voice bible saves you the time of editing a hundred mediocre drafts later.
Build this one in week two. The webinar recording and the rest of the Agent OS series live at nunncurtislabs.com/webinars.