AI Productivity Is a Leadership Problem (Not a Prompt Problem)

tl;dr

Most people treat AI like an unpaid intern. In reality, it’s a talented but directionless teammate. The difference between “meh” and “magic” isn’t the model—it’s how you lead: the clarity you give, the constraints you set, and the feedback loop you run.

AI marketing loves to say you’ve got “a PhD scholar in your pocket.” Maybe. But even a PhD needs a clear brief. AI mirrors your leadership. Treat it like a teammate, not a genie. 🤷🏻

Let’s start with two short stories to illustrate the point. 👇🏻

Two Short Stories (life → work)

Banana cake (life)

One evening my wife—she’s a doctor and currently vegan—asked for my banana bread recipe. I said, “Let’s ask AI.” She typed “How to make banana bread at home?” and got a standard egg‑and‑milk recipe. Then “without eggs”… now it had milk. Frustration level: rising. 😅

I stepped in with a tiny leadership move: give the context first.
“We’ve got overripe bananas, semolina (no AP flour), and dark chocolate chips. Exclude eggs and dairy. Any tried recipe with steps?”

AI returned a semolina banana cake that we actually baked—and it was good. The model didn’t suddenly get smarter; our brief did.

Founder pitch deck (work)

A week later I did the same thing at work. I needed a 10‑slide pitch outline.

Lazy ask → “Make me a pitch deck.” → Generic.
Leader brief → “Draft a 10‑slide deck for a B2B SaaS that automates inventory for small retailers. Audience: seed investors. Tone: confident, not hype. Must include: problem, solution, traction (3 pilots: Acme Mini‑Mart, FreshBazaar, KiranaKart), market size (use these numbers: …), pricing tiers, ask: $1M. Avoid buzzwords like ‘revolutionary’.”

The first draft was ~70% usable. My edits were surgical instead of rewriting from scratch. Time saved: real.

The Equation

$\text\{Output Quality\} = \text\{Model\} \times \text\{Context\} \times \text\{Feedback\} \times \text\{Taste\}$

Here’s how this equation flows in practice:

Model: The AI you’re using (ChatGPT/Claude/Gemini).
Context: Audience, goal, voice, constraints, sources.
Feedback: Tight loops; small nudges beat full rewrites.
Taste: Your editorial bar—what “good” looks like.

📌 Mantra: If “prompt engineering” is what to ask, context engineering is what the AI should already know before you ask.

Clarify the Why → outcome > activity.
Constrain the How → structure, length, tone, must‑include/avoid.
Close the Loop → fast feedback cycles; keep what worked.

Side‑by‑Side: Lazy Ask vs. Leader Brief

Vague ask:

Write a job description for a product manager.

Leader brief (context + constraints):

Write a 300-word job description for a **Remote AI Product Manager** role at a seed-stage startup. 
-------------------------------------
+ Audience: mid-level PMs who want to work in AI tools. 
+ Tone: transparent and human (avoid buzzwords). 
+ Must include: company vision (AI productivity tools), responsibilities (end-to-end product ownership), growth opportunity (early hire shaping culture). 

*Avoid generic lines like “rockstar PM” or “fast-paced environment.”*

The Leadership Loop: Diverge → Converge → Critique

Diverge: Ask for options first (5–10 angles/titles/outlines).
Converge: Combine the best parts (“Merge #2’s tone with #5’s structure”).
Critique: Have the model self‑grade against your acceptance criteria; then you give surgical feedback (keep A/B, cut C, add D).

Paste‑ready one‑liners:

Diverge: “Give me 7 title options and 3 opening hooks. Distinct angles.”
Converge: “Combine Hook #2’s tone with Title #5 and Outline #3. Draft 850±50 words.”
Critique: “Self‑grade for originality, specificity, one framework, one story, actionable checklist, clear CTA. Propose 3 fixes.”

The 5‑Minute Leader Brief (Copy‑Paste)

Use this as a reusable context pack before any AI task. No JSON needed—just paste and fill.

Goal: [Outcome and change for the reader]
Audience: [Who they are and what they care about]
Voice: Conversational, slightly witty, practitioner. Mirror tone from dev.to/sarthology and medium.com/@sarthology (tone only).
Sources: [Bullets/links/notes to ground the piece]
Constraints: H2s + bullets, 800–900 words, 1 named framework, 1 work story + 1 life story, avoid jargon.
Acceptance: Original insight, practical checklist, side‑by‑side example, clear CTA.

A Quick “Leadership” Checklist (tight 5)

Outcome in one sentence
Audience defined
Constraints (length, structure, must‑include/avoid)
Two examples (work + life)
Feedback plan (what counts as v1 “good enough”)

Closing + CTA

AI isn’t your worker. It’s your mirror. If you’re a lazy leader, you’ll get lazier outputs. Guide it with clarity and constraints, and it will surprise you—and save you time.

CTA: Try this today: write a 5‑line leader brief for your next AI task, time‑box 10 minutes, and compare the result to your usual one‑liner. You’ll feel the difference.