How to Use ChatGPT to Work Twice as Fast

Last Tuesday I drafted a 1,200-word technical spec, responded to 14 emails, and summarized three vendor contracts — all before lunch. I didn’t clone myself. I just stopped treating ChatGPT like a search engine and started treating it like a tireless assistant who never complains about scope creep.

Here’s exactly how I do it, with the prompts I actually copy-paste every day.

Stop Asking, Start Briefing

The single biggest mistake I see is vague requests. «Write me an email about the meeting» gets you garbage. «Summarize what happened in 2 sentences» gets you gold.

ChatGPT isn’t psychic. It needs context the way any new hire does. The mental shift that changed everything for me was this: don’t ask it a question, give it a brief. Role + task + format + constraints. That’s the template.

My go-to structure:

You are a [role]. [Task]. Format: [output format]. Constraints: [length, tone, audience].

For example, when I need to write a post-incident summary for a non-technical manager:

«You are a senior infrastructure engineer writing for a non-technical audience. Summarize the following incident in plain English. Format: 3 short paragraphs — what happened, why, what we’re doing to prevent it. Max 200 words. Avoid jargon.»

Then I paste in the raw log or my rough notes. Done in 40 seconds.

The Email Triage System That Actually Works

I used to spend 45 minutes every morning on email. Now it’s 15. The trick is batching. I don’t ask ChatGPT to write one email. I dump five situations and ask for five drafts in one shot.

My Monday morning prompt:

«I need drafts for 3 emails. For each one, keep the tone professional but direct. No fluff. Each email max 4 sentences. Here are the situations: [1. declining a vendor demo, 2. following up on an overdue invoice, 3. asking a teammate for a status update].»

It returns all three, numbered and formatted. I scan, tweak names and specifics, send. The key is «no fluff» — without that constraint, ChatGPT defaults to filler phrases like «I hope this email finds you well,» which I physically cannot send with a straight face.

Turning Raw Notes Into Usable Documents

I take meeting notes the old-fashioned way — messy, fragmented, half in shorthand. After the call, I paste them into ChatGPT with this prompt:

«Convert these rough meeting notes into a structured summary. Include: key decisions, open action items (with owner and deadline if mentioned), and any unresolved questions. Use bullet points. Be concise.»

This alone saves me 20 minutes per meeting. It also catches things I glossed over while multitasking — ChatGPT will pull out «pending budget approval» from three lines of chaos and flag it as an open item.

The same approach works for reading dense documents. Instead of reading a 40-page vendor proposal, I paste the key sections and ask:

«Extract the main commercial terms, any red flags or unusual clauses, and give me 5 questions I should ask before signing.»

That last part — «give me 5 questions I should ask» — is underused. It turns a passive summary into an actionable next step.

Code and Scripts Without the Stack Overflow Rabbit Hole

I’m not a developer, but I write bash scripts and Python utilities regularly for infrastructure work. ChatGPT has cut my scripting time by at least 60%.

What works: being specific about the environment. Not just «write a Python script that sends alerts,» but:

«Write a Python 3.11 script that reads from a JSON file at /var/data/alerts.json, filters entries where severity == ‘critical’, and sends them via POST request to a webhook URL stored in an environment variable called WEBHOOK_URL. Include error handling for missing files and failed requests. Add inline comments.»

That specificity matters. Without it, you get a generic skeleton you’d spend 30 minutes adapting anyway.

And when something breaks, I paste the error directly:

«This script throws the following error: [error text]. Here’s the relevant code: [code]. What’s wrong and how do I fix it?»

It’s not perfect — sometimes it confidently gives me wrong fixes. But it’s right 80% of the time and narrows the problem fast even when it’s not.

The «Think With Me» Mode

One of the most underrated uses isn’t about output — it’s about thinking. When I’m stuck on a decision, I use ChatGPT as a sounding board. Not to get the answer, but to stress-test my thinking.

«I’m deciding between two approaches: [A] vs [B]. Here’s the context: [situation]. Play devil’s advocate on both options. What am I not considering?»

The «what am I not considering» line is key. It forces the model to go beyond validating your idea and actually challenge it. I’ve caught at least three bad decisions this way — not because ChatGPT knew better, but because framing it as a challenge surfaces blind spots.

Prompt Fragments I Keep in a Pinned Note

Rather than reinventing the wheel every time, I keep a pinned note in Apple Notes (Settings > Notes > Pin to Top) with my most-used fragments:

• Tone control: «Direct and confident. No filler phrases.»
• Length control: «Max [X] words. Cut anything that doesn’t add value.»
• Format control: «Use bullet points. No headers unless the content has 3+ distinct sections.»
• Critical mode: «Argue against this. What’s the strongest objection?»
• Simplify: «Rewrite this for a non-expert. No jargon. 8th-grade reading level.»

I mix and match these into almost every prompt. They’re modular.

What It Still Can’t Do Well

It’s not magic. ChatGPT hallucinates specifics — numbers, dates, citations. I never trust any factual claim it makes without checking. It also has no memory between sessions by default (unless you use the Memory feature in Settings > Personalization > Memory), so I don’t rely on it knowing my preferences unless I state them in the prompt.

And for anything legally or financially sensitive? It drafts, I review, a professional signs off. Full stop.

The mindset shift is simple: ChatGPT isn’t smarter than you. It’s faster at drafting, formatting, and pattern-matching than you are. Give it clear instructions, treat it like a contractor (not an oracle), and it earns its keep every single day.