Delegation Frameworks That Work

Delegation fails not because of bad assistants but because of bad systems. You can hire the most talented EA in the world, and they'll still underperform if you delegate with vague instructions and no structure. The frameworks below are what separates people who get 10x value from their EA from people who cancel after three months.

None of these frameworks are complicated. They are straightforward systems that take minimal time to implement but fundamentally change how effectively your EA operates. The challenge is not understanding them; it is actually doing them consistently instead of falling back into the habit of firing off half-baked tasks and hoping for the best.

The 70% Rule

If someone else can do a task at least 70% as well as you, delegate it. This is the single most useful mental model for delegation, and most people get it wrong by holding out for 100%.

Here's the math. Say you spend 2 hours a week scheduling meetings. Your EA might take 2.5 hours at first and occasionally book a suboptimal time. That's maybe 70–80% of your quality. But you got 2 hours back. After a month, they're at 95% because they've learned your preferences. The gap between your quality and theirs closes fast—but only if you let them start.

The 70% rule also helps you identify what not to delegate. Investor negotiations, key hiring decisions, product strategy—these require your full judgment and context. If someone else would do a materially worse job than you, keep it. For everything else, let go.

A practical way to apply this: when you catch yourself about to do a task, pause and ask whether your EA could do it at 70% of your quality. If yes, add it to their queue and move on. If no, do it yourself. Over time, the 70% threshold shifts upward as your EA improves, and tasks that seemed impossible to delegate become routine.

The hidden cost of not applying the 70% rule is that you stay trapped doing $50/hour work when you should be doing $500/hour work. Every hour you spend scheduling meetings or formatting spreadsheets is an hour you did not spend on strategy, sales, or product decisions that actually move your business.

SOPs: Your Delegation Infrastructure

Standard Operating Procedures sound corporate, but they're the backbone of good delegation. An SOP is just a written description of how to do a task. It doesn't need to be fancy. A Google Doc with numbered steps works.

For every recurring task you delegate, write an SOP that covers: when to do it (trigger or schedule), the step-by-step process, the expected output, who to contact if stuck, and common edge cases. A good SOP is specific enough that someone who has never done the task before could follow it and produce acceptable results.

The trick is to build SOPs as you delegate, not before. The first time you hand off a task, record yourself doing it or write rough instructions. After your EA does it once, ask them to update the SOP with anything that was unclear. After three rounds, you have a polished procedure that anyone could follow. This means your delegation doesn't break if your EA is out sick or if you switch services.

Here is what makes the difference between a useful SOP and a document nobody references: include the edge cases. The SOP for booking travel should not just say "book flights on Delta when possible." It should say "book Delta when the price difference is under $150 compared to other carriers. If Delta is more than $150 more expensive, book the cheapest option but confirm with me first. Always book window seats. If no window seats are available in economy, check the price difference for economy plus and send me both options." Edge cases are where most delegation breaks down, and documenting them in the SOP prevents repeated back-and-forth conversations.

Store all SOPs in one place your EA can access easily. A shared Notion workspace or a Google Drive folder works. Label them clearly: "SOP - Travel Booking," "SOP - Email Triage," "SOP - Expense Reports." When your EA finishes a task, they should be able to find the SOP for their next task within 10 seconds.

Loom Over Docs (When It Matters)

Some tasks are hard to explain in writing. Anything involving software workflows, nuanced judgment, or visual decision-making is faster to show than to describe. Record a 3–5 minute Loom video walking through the process. Your EA can pause, rewatch, and reference it later.

The best approach is layered: a Loom for the first handoff, then a written SOP your EA creates from the video. The video captures the nuance. The document becomes the day-to-day reference. Together they cover 95% of questions.

One tip: Don't over-polish your Looms. The goal is clarity, not production quality. Hit record, talk through the task, stop recording. Done. A rough Loom sent today is infinitely more useful than a polished one you never get around to recording. Your EA does not care about your presentation skills; they care about understanding the task.

Loom is particularly effective for tasks that involve navigating specific software. Explaining how to create a report in your CRM through text requires screenshots, arrows, and extensive descriptions. A Loom showing you click through the process takes 90 seconds to record and is immediately clear. Build a library of these recordings organized by task category. Over time, you create a visual training program that any future EA can use to get up to speed.

The Authority Matrix

For each category of work your EA handles, define one of three authority levels:

Level 1: Act independently. Your EA handles these tasks end to end without checking with you. Examples: scheduling meetings with known contacts, booking travel within established preferences, triaging email, processing expense reports. These are tasks with clear rules and low risk of a bad outcome.

Level 2: Recommend and wait. Your EA does the work, presents options or a recommendation, and waits for your approval before proceeding. Examples: booking travel that exceeds budget guidelines, scheduling meetings with investors or board members, drafting replies to sensitive emails. These are tasks where your judgment adds value but the prep work can be done without you.

Level 3: Research only. Your EA gathers information but takes no action. Examples: competitive research, vendor evaluations, gathering data for decisions you need to make personally. These are tasks where you need the inputs but the output is entirely your call.

A common failure mode is leaving this ambiguous. Your EA isn't sure if they should book the flight or wait for your sign-off. They wait. You wonder why nothing got done. Define the authority level upfront for each task category. Write it down in a shared document. Revisit it monthly and move tasks from Level 2 to Level 1 as your EA demonstrates good judgment.

Async Communication Patterns

Most EA relationships are async, especially with remote or overseas assistants. The communication system you set up matters more than the communication tool you choose.

Daily standup message. Your EA sends a brief end-of-day update: tasks completed, tasks in progress, open questions. You scan it in 60 seconds. This replaces most of the status meetings people default to. The format should be consistent every day so you can scan it quickly. Three sections: Done, In Progress, Questions. That is it.

Task queue in a shared tool. Use Asana, Notion, Todoist, or even a shared Google Doc. Every task goes in the queue with a deadline and priority level. Your EA works the queue. You add to it. Nobody has to remember what was said in a Slack thread three days ago. The task queue is the single source of truth for what needs to happen. If it is not in the queue, it does not exist.

Urgency signals. Define clear labels. "When you get a chance" means this week. "Today please" means today. "Urgent" means drop everything. If you use "urgent" more than once a week, nothing is urgent and your EA will stop treating it that way. Reserve "urgent" for genuine time-sensitive situations: a meeting in two hours needs materials, a flight needs rebooking due to a cancellation, an investor responded to a time-sensitive ask.

Weekly planning session. In addition to the daily standup, a 10-15 minute weekly sync (Monday morning works well) lets you align on priorities for the week, flag anything unusual coming up, and review what went well or poorly the previous week. This keeps the relationship on track without requiring daily meetings. After the first month, many EA relationships run almost entirely on the daily async standup and the weekly planning session, with ad-hoc messages only for genuinely new or time-sensitive items.

Putting It All Together

The playbook is straightforward. Use the 70% rule to decide what to delegate. Create a Loom for the first handoff. Have your EA write the SOP after doing it once. Define the authority level so they know how much autonomy they have. Communicate async with a daily standup and a shared task queue.

None of this is complicated. But most people skip it because it feels like overhead. It's not. It's the difference between an EA who's an expense and an EA who's an investment. Build the system once, and it pays dividends for as long as you have an assistant. These frameworks also transfer. If you switch EA services or your assistant changes, the SOPs, authority matrix, and communication patterns carry over. Your next EA gets up to speed in days instead of weeks because the infrastructure already exists.