Skip to main content

Turn Your AI Agent into a Proactive Assistant

Most people use AI the same way: open a chat, type a question, get an answer, close the tab. That's reactive AI. This playbook shows you how to shift to proactive AI — an assistant that knows your preferences, remembers your recurring tasks, and works ahead of you.

info

This content was developed with AI assistance and is regularly reviewed for accuracy.

What You'll Accomplish​

By the end of this playbook, you'll have:

  • A persistent context file your AI can reference across conversations
  • A recurring task list your AI can execute on a schedule
  • A "heartbeat" system that keeps your AI working between your sessions
  • A framework for expanding your assistant's responsibilities over time

Time: 45-60 minutes for initial setup, then 10 minutes weekly to maintain

Difficulty: Intermediate — you should be comfortable with basic AI conversation and have a prompt library started.

Prerequisites​

  • An AI platform you use regularly (ChatGPT, Claude, Gemini, or an agent framework like LangGraph or CrewAI)
  • A text editor or notes app for writing context files
  • 3-5 recurring tasks you currently do manually

The Playbook​

Step 1: Identify Your Repetitive Patterns​

Goal: Find the tasks your AI assistant should handle without being asked.

Action: Write down everything you do repeatedly — daily, weekly, or monthly. Focus on tasks that follow a predictable pattern.

Examples of recurring tasks:

  • Morning briefing: summarize overnight emails and prioritize the day
  • Weekly report: gather metrics from multiple sources, write a summary
  • Meeting prep: pull context on attendees, review past notes, draft an agenda
  • Content scheduling: check what's due, draft social posts, queue newsletters
  • Inbox triage: categorize messages by urgency and suggest responses

Checkpoint: You should have at least 5 recurring tasks written down. Star the 2-3 that eat the most time.

Step 2: Write a Persistent Context File​

Goal: Give your AI the background knowledge it needs to work proactively.

Action: Create a document called MY_CONTEXT.md (or similar) that captures who you are, what you care about, and how you like things done.

Template:

# About Me
- Role: [Your job title and what you actually do day-to-day]
- Team: [Who you work with, their roles]
- Tools: [Software and platforms you use regularly]

# Preferences
- Communication style: [Formal/casual, brief/detailed]
- Report format: [Bullet points, narrative, specific templates]
- Priorities: [What matters most this quarter]

# Recurring Tasks
- Daily: [List tasks]
- Weekly: [List tasks]
- Monthly: [List tasks]

# Don'ts
- [Things the AI should never do or assume]

Why this matters: Without persistent context, every conversation starts from zero. With it, your AI can skip the "let me explain my situation" preamble and get straight to useful work.

Checkpoint: Your context file is at least 15-20 lines and covers your role, preferences, and top recurring tasks.

Step 3: Build Your First Recurring Task Prompt​

Goal: Turn one recurring task into a repeatable AI workflow.

Action: Pick your most time-consuming recurring task from Step 1. Write a prompt that:

  1. References your context file
  2. Specifies what input the AI needs (you'll provide this each time)
  3. Defines the exact output format
  4. Includes quality criteria

Example — Weekly Report:

Context: Refer to MY_CONTEXT.md for my role, team, and report preferences.

Task: Generate my weekly status report.

Input I'll provide:
- Raw notes from this week's work
- Key metrics (I'll paste numbers)
- Any blockers or decisions needed

Output format:
1. Executive summary (2-3 sentences)
2. Completed this week (bullet points)
3. In progress (bullet points with % complete)
4. Blockers requiring attention
5. Next week's priorities

Quality criteria:
- Use active voice
- Quantify results where possible
- Flag anything that's been blocked for more than 1 week

Checkpoint: You have one fully written recurring task prompt that references your context file.

Step 4: Create a Heartbeat System​

Goal: Set up a routine where your AI assistant checks in and works proactively.

Action: A "heartbeat" is a scheduled prompt that kicks off your AI's proactive work. Instead of waiting for you to ask, you trigger a standard workflow at set intervals.

Daily heartbeat example:

Good morning. Here's what I need you to do:

1. Review today's calendar (I'll paste it below)
2. Flag any meetings that need prep
3. Check my task list and remind me what's due today
4. Draft a priority list for today based on deadlines and importance

Calendar:
[Paste today's schedule]

Task list:
[Paste current tasks]

Weekly heartbeat example:

It's Monday morning. Weekly review time:

1. Summarize what I accomplished last week (from the notes below)
2. Identify any tasks that carried over and should be escalated
3. Review this week's calendar and flag prep needed
4. Suggest my top 3 priorities for the week

Last week's notes:
[Paste notes]

This week's calendar:
[Paste calendar]

For agent frameworks: If you're using a framework like LangGraph or n8n that supports scheduled tasks, you can automate the heartbeat entirely — the agent runs the prompt on a cron schedule and delivers results without you lifting a finger.

Checkpoint: You have at least one heartbeat prompt written and have run it once successfully.

Step 5: Expand and Iterate​

Goal: Grow your assistant's responsibilities over time.

Action: Once your first recurring task and heartbeat are working, expand gradually:

  1. Add one new task per week. Don't try to automate everything at once.
  2. Update your context file as your role, tools, or preferences change.
  3. Create task-specific context files for complex workflows (e.g., a separate file for meeting prep that includes your meeting notes template).
  4. Track what works. Keep a simple log of which prompts save you the most time.

Expansion path:

  • Week 1: Daily heartbeat + one recurring task
  • Week 2: Add meeting prep workflow
  • Week 3: Add weekly report automation
  • Week 4: Add content or communication workflows
  • Month 2: Review and optimize — cut prompts that aren't saving time, double down on ones that are

Checkpoint: After one month, you should have 3-5 recurring workflows running through your AI assistant.

Common Pitfalls​

PitfallFix
Context file becomes outdatedSchedule a monthly review — add it to your heartbeat
Trying to automate everything at onceStart with one task, prove it works, then expand
Forgetting to provide fresh inputBuild a checklist of what to paste into each heartbeat
AI output quality drifts over timePeriodically review outputs against your quality criteria
Overcomplicating the systemIf a task takes more effort to automate than to do manually, skip it

Key Takeaways​

  • Reactive AI waits for questions. Proactive AI anticipates work. The shift happens when you give your AI persistent context and recurring instructions.
  • Context files are the foundation. The better your AI knows your preferences and situation, the less you have to explain each time.
  • Heartbeats create rhythm. Scheduled prompts turn your AI from an on-demand tool into a working assistant.
  • Start small and expand. One well-automated task beats ten half-built workflows.

Next Steps​

SponsoredYour product hereReach people actively learning AI — flat monthly rates, no ad networks