The Automated Manager

Your job as a middle manager isn’t to write the perfect email. Your job is to get decisions made with minimum risk and minimum mess. In regulated companies (or any place where people forward messages to legal, risk, and compliance), writing gets slow because every draft has to survive scrutiny. If you draft “open-ended,” you create more edits. If you draft “loosely,” you force reviewers to guess what you meant. If you draft “too polished,” you force reviewers to dig for what changed and why.

The EIC Loop (Draft→Edit→Approve) solves this by making AI output something you can safely edit. You don’t ask the AI to be smart and guess policy. You ask it to draft to your rules, then stop. Then you run a repeatable edit pass that checks structure, risk, and clarity in the same order every time.

Here’s the core idea: AI should produce a workable first draft that matches corporate constraints - tone, length, formatting, and what not to include. You should own the final accuracy and the “editor decisions” (what stays, what moves, what gets deleted). That reduces friction because reviewers see consistent formatting and consistent risk handling, so they don’t burn time re-parsing your message.

Tanya (39, Operations Manager at a regulated fintech) runs into this every week. She’s not short on content - she’s short on time. She spends her mornings chasing thread updates and her afternoons rewriting decks after feedback. When she switches to EIC Loop (Draft→Edit→Approve), she stops treating AI like a writer and starts using it like a draft engine she can reliably edit.

Practical takeaway: You don’t need “better writing.” You need a workflow that makes each draft reviewable faster and makes your edits repeatable.

How to run the EIC Loop without triggering rework

You’ll use the EIC Loop (Draft→Edit→Approve) every time you need a corporate document: an email, an executive update, a compliance note, or a slide narrative. The loop keeps AI output predictable, which matters in strict environments.

1. Draft (AI writes to your rules) Use a prompt that locks the role, scope, constraints, and formatting. Then force the AI to generate a draft that is “review-ready,” not “final-final.” Copy-paste prompt template: > Role: You are my Draft Assistant for internal corporate communication. > Task: Draft ONLY the requested document based on the inputs I provide. > Constraints: > - Use professional, neutral tone. > - Do not invent facts or claim approvals. > - If any required detail is missing, add a section titled “Missing Info” with bullet points. > Output format: > - Subject line (if email) > - 3-6 short paragraphs max > - Bullets only when listing items > - End with “Proposed next step:” and one sentence > Inputs: [paste thread notes / meeting bullets / facts] > Draft now.

2. Edit (You run a checklist in a fixed order) You don’t “read it and react.” You edit in passes. First pass fixes structure. Second pass fixes risk language. Third pass fixes clarity and length. This reduces the chance you’ll miss something and cause another review cycle.

Editor-in-Chief edit checklist (run in this order): - Structure: Do the first 3 lines answer “what changed” and “what you need”? - Claims: Did the draft avoid “we confirmed” unless you truly have confirmation? - Compliance-safe language: Did it avoid absolute promises (like “will comply”) when you only have “we plan to” or “we are reviewing”? - Missing Info: Did you fill the gaps or remove the parts that depend on them?

3. Approve (You stamp it and send) Approval means you take responsibility: you confirm the facts, you confirm the tone, and you confirm the next step. In regulated settings, that last step prevents “someone else will handle it” messages that create loops.

4. Log (one line: what caused the edit time) After sending, write a single note: “Edit driver: tone too casual / missing policy reference / unclear ask.” This becomes your next prompt improvement.

Concrete example (Tanya’s weekly rhythm): She drafts an executive update from her meeting notes. Then she edits using the checklist. Finally she sends with a clear “Proposed next step” that tells readers exactly what to do next and by when. Her reviewers stop replying with “What do you want from us?” because she forces that at the end.

Practical takeaway: Your loop reduces friction because the AI draft always lands in the same shape, and your edits always check the same risk points.

Putting the EIC Loop into practice: Tanya’s deck-ready update flow

Let’s run a realistic scenario that matches what you actually deal with: you need an executive update from messy notes, and legal/compliance tends to request changes after the first pass.

Scenario Tanya has a Monday stakeholder meeting. She has: - A rough set of bullet points - A few email fragments from the thread - One compliance comment that she remembers, but doesn’t want to misstate

She needs a short update by end of day, formatted so leadership can scan it and compliance can see risk language clearly.

Step-by-step EIC Loop execution

1. Collect inputs in one place Paste your raw bullets and any quotes into a single doc. Include dates and owners for each bullet. Expected outcome: The AI draft doesn’t waste time asking questions you already have.

2. Run the Draft step with strict formatting Paste this template:

> Role: You are my Draft Assistant. > Task: Draft an executive update for leadership based on my inputs. > Constraints: > - No invented facts. > - Use only what I provide. > - If something depends on an open approval, label it “Pending approval” and explain what’s pending. > - Keep it under 220 words. > Output format: > - Title (one line) > - “What changed:” (2 bullets max) > - “Current status:” (3 bullets max) > - “Risks / compliance notes:” (bullets only; no legal conclusions) > - “Proposed next step:” (one sentence with owner + timing) > Inputs: [paste Tanya’s bullets + compliance comment text] > Draft now.

Expected outcome: You get a draft that already matches leadership’s scan pattern and avoids legal overreach.

3. Run the Edit step using a fixed pass order Edit pass 1 (structure): Make sure “What changed” comes first and reads like a summary, not a diary. Edit pass 2 (claims): Replace any “confirmed” with “reported” or “under review” if you didn’t verify. Edit pass 3 (risk language): Remove absolutes. Use “we are reviewing,” “we plan to,” or “we need guidance on” when appropriate. Edit pass 4 (length): Cut anything that doesn’t support a decision.

Expected outcome: Fewer “rewrite” requests because the draft already reflects compliance-safe phrasing.

4. Approve and add one decision trigger Tanya adds a single line in the “Proposed next step” sentence that tells the reader what action to take.

Expected outcome: You reduce the email thread churn because you stop sending vague updates.

Quick checklist (do this every time) - Draft: AI output uses your exact format and ends with “Proposed next step.” - Edit pass 1: First lines answer “what changed” and “what you need.” - Edit pass 2: No invented facts; no false certainty. - Edit pass 3: Compliance-safe language only; label anything open as “Pending approval.” - Approve: You verify owners, timing, and claims before you hit send.

Practical takeaway: When you force draft formatting and run edits in passes, you stop “re-drafting” every time someone asks for clarity.

What to watch for (and how to fix it fast)

Mistake: AI drifts into “final” language If the draft sounds like a signed-off policy document, reviewers will treat it like a commitment and push back harder.

Do this: Add a constraint that forbids legal conclusions and forces “Pending approval” labeling when needed. Not this: Ask for “final compliance wording” in the Draft step.

Mistake: You skip the edit pass order If you edit by vibes, you’ll miss one risk phrase and create a second review cycle.

Do this: Run structure → claims → risk language → length in that order every time. Not this: Fix tone after you already changed facts; that forces you to re-check claims again.

Mistake: You copy-paste the same prompt without updating constraints Regulated environments change what’s safe to say. If your prompt still allows “confirmed,” you’ll keep generating drafts with that problem.

Do this: Update the Draft prompt constraints for the specific document type (email vs. deck update vs. compliance note) and the current approval state. Not this: Reuse a generic prompt for everything when you know some messages require “Pending approval” language.

Practical takeaway: Friction usually comes from one mismatch: AI draft certainty vs. real approval state. Fix that mismatch first.

If you want fewer edits tomorrow, start by running the EIC Loop (Draft→Edit→Approve) today with the strict Draft template and the fixed edit pass order. That rhythm builds a workflow your reviewers can trust - and it sets you up to automate the next category of manager writing without losing control.

End of chapter one. 8 more chapters and about 16,813 words in the full book.