PHASE 01 · MINDSET

The Shift From Executor to Director

Before you write an SOP or open an AI tool, separate your job title from the workflows you own. This phase turns anxiety and hype into a task map you can investigate.

⏱ 30 min read · 1 exercise Mindset Don't skip

The planning problem

No defensible countdown predicts when an individual job will change or disappear. AI exposure varies by task, industry, adoption, economics, regulation, data access, and the quality standard required.

A better planning unit is the task. Design tweaks, copy iterations, operations checklists, code scaffolding, and basic analysis may each have different assistability, failure costs, and review needs—even inside the same role.

Treat this as a portfolio question: which tasks should remain manual, which deserve an assisted experiment, and which could become a governed workflow after evidence?

★ THE CORE INSIGHT

Predictable steps are often easier to test for assistance, but predictability alone does not establish that automation is useful or safe. Data quality, exception frequency, consequence, cost, and review capacity matter too.

Your responsibility is to make those factors explicit and keep a named human accountable for the outcome.

Three valid approaches

The right approach can differ by task. Use evidence and consequence—not identity or urgency—to decide.

Approach A · Ad hoc assistance

Use AI for isolated drafts or exploration. This may be appropriate for low-consequence work, but results are difficult to reproduce unless instructions, inputs, review, and failures are recorded.

Useful, but hard to govern
Approach B · Governed workflow ownership

Define the task, acceptance criteria, data boundary, tests, approval point, and rollback. Use an assistant only where measured performance and consequence justify it.

This roadmap teaches this approach
Approach C · Keep it manual

Retain human execution when the evidence is weak, the task is rare, the data is restricted, failures are consequential, or review costs erase the benefit. Revisit only when conditions change.

A valid control decision

The accountable workflow-owner archetype

Picture a manager who makes ownership, instructions, evidence, and review boundaries clear. Three practices transfer well:

  1. They could articulate any task in their function in enough detail that a competent stranger could deliver it. They had the recipe in their head.
  2. They picked the right person for each task instead of grabbing the nearest available body. They knew whose hands were good for which thing.
  3. They reviewed work fast and corrected sharply — they didn't redo it themselves, they returned it with the exact note that prevented the same mistake next time.

AI assistants are not employees and the analogy has limits: they do not hold responsibility, understand context reliably, or own consequences. You remain accountable for scope, access, review, and action.

Imagine a four-brand DTC team testing one caption workflow. The team documents the brief, classifies brand and customer data, creates a representative test set, defines prohibited claims, and keeps publishing with a human reviewer.

The useful evidence is not headcount, speed, or margin assumed in advance. It is whether the workflow meets its factuality, voice, safety, and review-time criteria on repeatable tests.

Why task-level evidence matters

Experts can overestimate or underestimate what a system can do. Both errors are costly.

Do not infer a job outcome from a demo or a percentage of tasks. Measure representative cases, important edge cases, review time, failure severity, and the economic value of the complete workflow.

The trap is making a blanket claim in either direction: "AI cannot affect my work" or "AI can replace almost all of it."

Break the workflow into tasks, test them, and retain human execution wherever evidence, accountability, or consequence requires it. This roadmap does not predict pricing or market demand.

Habits that support responsible adoption

These habits make workflow experiments easier to evaluate and govern. They are not a readiness score or a prediction of career success.

01They write things down obsessively. Their workflows live in documents, not in their head. This makes them naturally SOP-ready.
02They are unattached to "doing it themselves." Their identity is the outcome, not the keyboard. Delegating to AI doesn't trigger an ego defense.
03They are good at giving feedback in writing. Vague managers raise vague employees — and vague AI workers. Sharp written feedback is the training signal.
04They make failures observable. They test representative and adversarial cases, record errors, and rerun regressions after every material change.
05They design for change. They version instructions, assign owners, set review dates, and define when to pause or retire a workflow.

If several habits are missing, choose one to practice before expanding access or automation. A checklist count does not establish readiness.

The self-assessment

Be brutally honest. There's no score to share with anyone. The point is to know where you actually start.

Where Are You Right Now?

Do not total the boxes into a score. Use unchecked items as preparation gaps. Start Phase 2 with one low-consequence workflow, synthetic or public data, and no write access.

The exercise — your commitment moment

This short exercise creates the first artifact you will revise as evidence changes. It does not predict completion or success.

Open a new doc. Title it: AI Director — My Roadmap.md

  1. Write today's date.
  2. Write your role in one sentence. (Not your title — your actual job. "I help DTC brands run paid social on Meta and TikTok.")
  3. Write the 3 things you spend the most time on in a typical week.
  4. Next to each one, write "D" if it's mostly delegate-able (predictable, repeatable, has a clear recipe), "J" if it's mostly judgment (case-by-case, no two are alike), or "M" if it's mixed.
  5. At the bottom, set a review date: "By [date], I will evaluate one low-consequence task against documented acceptance criteria and decide whether to keep it manual, assist it, revise it, or retire the experiment."

That's the commitment. Save the file. You'll add to it at the end of every phase.

★ KEEP THIS NEAR YOU

From this moment until the end of the roadmap, every time you're about to do a task, ask yourself one question: "Is this a task I should be DOING, or a task I should be DIRECTING?"

The answer can remain "doing" when evidence, consequence, or accountability demands it. The goal is an explicit decision, not a target automation percentage.

Recap

Onward. Phase 2 is where you stop talking and start mapping the actual job.