Most winery leaders run their inbox the way a tasting bar runs on a busy Saturday — they triage by volume instead of importance. The wholesale buyer's email that needed a 3-line reply sits at the bottom because it is not urgent. The compliance notice gets a 20-minute response that could have been delegated. The vendor follow-up nobody chased turns into a six-month invoicing dispute. By 4 PM, you have answered emails for two hours and still have 47 unread.
The Executive Assistant agent fixes the part of that problem software can fix. It reads your inbox every hour during the working day, drafts replies for the routine ones, flags the urgent ones, and writes a 5-line pre-meeting brief 30 minutes before every external meeting on your calendar. You walk into every meeting with the last conversation already in your head. Your inbox never holds something you needed to see four hours ago.
This is the per-agent breakdown for the Executive Assistant in the Agent OS Webinar series. Week four of the 90-day install order. It comes after the Researcher, Speechwriter, and Personal CFO because by week four you have built enough trust in the pattern that handing the agent your inbox feels reasonable.
What the Executive Assistant reads (Inputs)
The Executive Assistant is the agent with the largest live-data surface. Five inputs cover what it needs.
- Your inbox. New mail since the last run. The agent reads each message, identifies who it is from, and figures out what a useful reply would look like.
- Your calendar. Today and the week ahead. The agent knows which meetings are coming and who you are meeting with, and remembers what was discussed last time.
- Your contact and people notes. Whatever you have written about each person — their role at their company, the projects you share, the last time you talked, their preferred channel. The Executive Assistant is only useful in proportion to the quality of these notes.
- Your prior threads with each person. When a wholesale buyer emails, the agent reads the last four threads with that buyer before drafting a reply. The reply remembers what you committed to and where you left off.
- Your voice signal. A short voice document or a pointer to your Speechwriter's voice bible. The Executive Assistant uses this to draft replies that sound like you, in the channel you are replying through.
The inputs overlap with the Researcher's, but the use is different. The Researcher reads context to brief you. The Executive Assistant reads context to draft for you.
What it does with that material (Skills)
Four skills, in rough order of how much time each one saves.
- Hourly inbox triage. Every hour during the working day, read new mail. Sort into three buckets: drafts ready for your review, flags that need your attention, archive-safe items the agent files away. The bucket distribution lets you walk into the inbox knowing what is there before you open it.
- Auto-draft replies in your voice. For the routine emails — the wholesale order confirmation, the vendor scheduling, and similar predictable cases — the agent writes a draft and queues it. You open the draft, change a word if you want, and hit send. The act of replying is reduced to a few seconds of approval.
- Pre-meeting briefing. Thirty minutes before any external meeting on your calendar, the agent writes a 5-line brief into your daily note. Who the meeting is with, what was last discussed, what is open, what they care about. You walk in with the last conversation already in your head.
- Calendar protection. When someone asks for a meeting at a time that would compromise a block you set aside for focused work, the agent flags the conflict and proposes alternatives. It does not move things on your calendar without you. The flag goes in your daily note, with two or three suggested times to offer back.
The Executive Assistant does not press send. The hard rail again: the agent drafts and you send. The agent has not yet earned the right to speak in your voice without you watching. Some leaders never give it that right, and that is fine.
The schedule (Scheduled Tasks)
Three cadences.
- Hourly inbox check, 8 AM to 6 PM on weekdays. New mail comes in. The agent triages and drafts. Drafts sit in Gmail's drafts folder, addressed but unsent. Flags land in your daily note.
- 30 minutes before any external meeting on the calendar. The pre-meeting brief writes itself into your daily note next to the calendar entry. You read it on the way into the meeting.
- On-demand email triage. When you want to clear the inbox on a Sunday evening, the agent walks the whole inbox start to finish, surfaces what is worth your attention, drafts replies for the rest. The session ends with the inbox triaged.
The hourly cadence is the one that changes how it feels to run your inbox. The pile never gets bigger than one hour's worth of mail. You scan the new drafts during a 90-second break between meetings, approve the ones that are clean, kick back the ones that miss the mark.
What lands on your desk (Outputs)
Three outputs, all in places you already look.
- Gmail drafts. The drafts sit in your drafts folder, ready to send. The agent never sends. You do.
- Daily-note flags. The urgent items and the ambiguous ones land in your daily note in your vault. You see them when you open your morning briefing, alongside everything else.
- Pre-meeting briefs. Five lines per meeting, in the daily note, 30 minutes before the meeting. Each brief includes the last four touchpoints and any open items the other side cares about.
Nothing autonomous leaves your account. The agent reads, drafts, flags. The signing-off is yours.
The goal: walking in informed, inbox under control
The Executive Assistant exists for two outcomes that connect. You walk into every meeting knowing what was last said. Your inbox never holds anything you needed to see four hours ago. Both outcomes are achievable with a 5-line brief and an hourly triage cycle. Neither was achievable when the inbox ran you instead of the other way around.
The second-order effect is that the rest of your day opens up. The 90 minutes you used to spend on inbox triage shrinks to 15 minutes of reviewing drafts. The 20-minute pre-meeting scramble shrinks to 30 seconds of reading the brief. Across a five-day week, the agent gives you back somewhere between four and seven hours of working time. Those hours go to the work only you can do — the conversations and the strategy calls where the inbox is the wrong vehicle.
A worked example at Not Really Wines
At Not Really Wines, our fictional demo winery at NunnCurtis Labs, Maren Holloway, Co-Founder and CEO, has a tasting room walkthrough scheduled for Tuesday at 10 AM with Tessa Brennan, who runs hospitality at the Tasting Bar in downtown Healdsburg. NRW makes premium non-alcoholic wine, vacuum-distilled from Sonoma fruit. The walkthroughs are weekly. The punch list is always long.
At 9:30 AM Tuesday, the Executive Assistant writes into Maren's daily note:
10:00 AM — Tasting Bar walkthrough with Tessa. Last four emails: (1) Tessa proposed two new flight formats for The Society subscribers visiting — Maren agreed in principle but said "let's walk through pricing Tuesday." (2) Tessa flagged a temp staffing gap for the May holiday weekend — Maren said she would call Sam in production to see if anyone could float over. (3) A Yelp review last Friday mentioned slow service during the lunch rush — Tessa was already drafting a response. (4) Tessa asked about reordering the Pinot Noir vintage for the bar pour — Maren said yes, ask Marcus for the budget. Open punch list: pricing on the new flights, Sam-or-temp coverage decision, response to the Yelp review, reorder approved.
Maren reads four lines on the way to the floor at 9:55. She walks in already remembering where they left off. The walkthrough takes 45 minutes instead of 70 because they do not spend the first 25 minutes reconstructing context. Tessa noticed by the third week that Maren came in better prepared. By the fourth week, Tessa started bringing her own version of the brief, written in the same shape. The pattern propagated.
Why this is week four of the install order
The Executive Assistant comes fourth for one reason: the inbox is where the most context lives, and the most context means the most things to get wrong. Installing this agent first risks an early bad draft that costs trust. Installing it after the Researcher, Speechwriter, and Personal CFO means you have seen the pattern work in three lower-stakes domains. By week four you trust the read-and-draft model. The inbox is just the next domain.
Two more reasons:
- The Executive Assistant uses outputs from the agents you already have. The voice bible the Speechwriter reads is the same one the Executive Assistant uses to draft replies. The contact notes the Researcher pulls from before a meeting are the same ones the Executive Assistant reads before drafting. You are not starting from scratch on the inputs.
- The compounding effect is faster here than anywhere else in the chain. By the end of week four, the Executive Assistant has read 200 to 400 of your emails and 80 to 120 of your drafts and replies. The drafts it produces in week six are noticeably closer to your voice than the ones in week four. The agent improves on its own as your archive grows.
Build this one in week four. The webinar recording and the rest of the Agent OS series live at nunncurtislabs.com/webinars.