Workflow 001: The Morning Brief Agent

Every morning at 8am, an AI texts me my orders for the day.

His name is Albus. He reads my calendar, my email, my CRM, and my task manager, then sends me one iMessage that tells me what actually matters. Not everything. What matters.

A morning brief agent is a scheduled AI task that reads your calendar, inbox, CRM, and task manager each morning, then sends one short message ranking the day against a goal you set. You build it by connecting those four sources to an AI assistant, saving a prompt that defines the brief's structure, and scheduling it to run before you start work. It takes about 30 minutes to set up and replaces the twenty minute audit most founders run by opening five apps in sequence.

Here's the entire workflow so you can build your own.

What this replaces

The first hour of my morning used to be an audit. Calendar, then inbox, then Notion, then back to the inbox to double check. Five apps just to figure out where the day's risk was.

The worst part was never the time. It was starting every day in reactive mode, letting whoever emailed loudest decide my priorities.

Now it's one text. Sixty seconds. Then I start the day already knowing where the leaks are.

Blueprint

The process in plain language:

Every weekday at 8am, an agent pulls four sources: today's calendar, my inbox (already cleaned by my email agent an hour earlier), my CRM, and my task manager. It sorts everything against one filter, my current 90-day goal. Then it sends a single message with five sections: the calendar ranked by weight, quick wins under 2 minutes, focus blocks that need real time, rescued tasks, and its own read of the day.

Trigger: the clock, every weekday morning. Inputs: calendar, inbox, CRM, task manager, my 90-day goal. Output: one iMessage I can read on my phone before I get out of bed.

One design note before you build: I did not start with the AI. I started by watching myself for a few days. Which apps did I actually open every morning? What was I looking for in each one? The brief covers those and nothing else. Most people build the brief they imagine wanting, then abandon it, because it answers questions they never actually ask.

Organize

Before the agent existed, the information had to live somewhere it could reach.

• Calendar: every call, buffer, and personal block actually on the calendar. If it's not on the calendar, Albus can't protect it.
• Email: one inbox. The agent looks for open loops: unsigned documents, unanswered proposals, people waiting on me.
• Tasks: one task manager, one place. Mine is Notion.
• CRM: a client database the agent can read, so when it sees a meeting it knows who the person is, not just a name on a calendar block.
• The goal: written down where the agent can see it. This one line changes everything, because "important" is meaningless without something to measure it against.

This is the unglamorous truth of every AI workflow. The agent is only as good as the system underneath it. Systems before AI.

Systematize

The agent runs as a scheduled task in Claude. A scheduled task is an instruction the assistant runs on its own timetable, with no one in the chat. I use Cowork with my calendar, Gmail, and Notion connected.

Here is the sanitized version of the prompt. Fill in the brackets and it runs.

You are [AGENT NAME], my morning brief agent. Every weekday at [TIME],
pull from my calendar, email, and task manager, then send me one message
that tells me what actually matters today.

My context:
- Name: [YOUR NAME]
- Timezone: [YOUR TIMEZONE]
- Current 90-day goal: [YOUR GOAL, ONE LINE]
- My tools: [e.g. Google Calendar, Gmail, Notion]

Build the brief in this exact structure:

1. HEADER: today's date plus my 90-day goal. Every brief starts by
   reminding me what I'm building toward.

2. CALENDAR: today's events sorted by weight, not just time.
   - Heavy meetings (decisions, money, external people): mark with ⛰️
   - Routine syncs: mark with 🪨
   - Buffers and personal blocks: mark with 🪶
   - Flag anything unusual: a key attendee declined, an overlap, a
     meeting with no agenda. Tell me what to confirm before I waste
     the slot.
   - Close the section with my free blocks and the total call count.
     Be honest when the day is brutal.

3. QUICK WINS: 2 or 3 things I can clear in under 2 minutes, pulled
   from email and tasks. Unsigned documents, unanswered proposals,
   replies that unblock other people.

4. FOCUS BLOCKS: 1 or 2 things that need real time. Give me the
   backstory in one line: how long it's been open, who's waiting,
   why today.

5. TASKS: anything overdue or rescued from my task manager. If a tool
   connection fails, say so plainly and tell me how to fix it. Never
   fake data.

6. NOTE: close with your read of the day. Where the risk is, what to
   protect, what goes cold if I ignore it. Talk to me like a chief of
   staff, not like a calendar app.

Rules:
- Prioritize against my 90-day goal, not against urgency alone.
- Short lines. No fluff. I read this on my phone in 60 seconds.
- If something looks wrong, lead with it.
- You are allowed to have opinions. That is why you exist.

How do you set it up?

1. Open Claude and connect your calendar, email, task manager, and CRM through their connectors.
2. Save the prompt as a skill with your details filled in. Naming it is optional but recommended. Mine is Albus, the coordinated productivity wizard.
3. Run it manually a few times before scheduling anything. Fix what feels off. The first brief is never quite right, and that's normal.
4. Once the format feels like yours, set it as a recurring scheduled task every weekday morning, delivered wherever you actually look. Mine arrives by iMessage.
5. When your business changes, tell it: "let's update your skill file." The brief gets better every month you use it.

Assign

This is the section most workflow guides skip, and it's the most important one.

What the agent does alone: reading the sources, ranking the day, spotting the anomalies, writing the brief. No human touches any of it.

What stays human: every decision. Albus told me a meeting might be dead because a key attendee declined. He did not cancel it. He told me a warm lead finally replied after three follow-ups. He did not write the reply. He recommends. I decide.

What if it gets something wrong?

That line is deliberate. The morning brief is a read-only workflow: it looks at everything and touches nothing. That's what makes it safe to run unattended while I'm asleep. A wrong call costs me a bad recommendation I can ignore in two seconds. When an agent both decides and acts, you need a much tighter leash. When it only observes and reports, you can let it run.

Rule of thumb from this build: give AI the audit, keep the verdict.

Integrate

Albus runs at 8am on purpose, not 7.

My email agent runs at 7am. She cleans the inbox first: junk out, everything categorized, drafts prepared. Albus reads the result an hour later. So the brief never wastes a line on mail that has nothing to do with me. Two agents, sequenced like a relay, and the order is the design.

That's what integration actually means. Not "connected to my tools" but agents that account for each other:

• What Albus flags as a focus block often becomes the first work session of the day.
• The open loops he catches (unsigned agreements, cold leads) feed straight into follow-up workflows.
• When a tool connection breaks, he reports it in the brief itself, which means the system tells me when the system is broken.

One agent, one message, one honest read of the day. That's the whole workflow.

Tomorrow: The Me Skill, the file that teaches AI who I am, so every other workflow sounds like me and knows my context. It's the foundation every agent in my stack is built on.