A few weeks ago, one of my students from the Allianz team pulled me aside after class and showed me something on her phone. It was a time log — a simple spreadsheet she’d been keeping for the past month. Before joining the AI In Action Bootcamp, she was logging 60+ hours a week. Not 60 hours of meetings or strategy. Sixty hours of work that included a huge amount of admin, writing, report prep, email, and research — the scaffolding around her actual job.

After four weeks in the bootcamp, she was at under 15 hours for that same category of tasks.

She wasn’t working less. She was doing the same job. She just wasn’t carrying the same weight around it anymore.

That moment stuck with me, because it captures exactly what I’m trying to teach — and exactly what most people get wrong about AI.

AI Doesn’t Replace Thinking. It Removes the Friction Around Thinking.

Here’s the thing about professional work: the part that actually requires you — your judgment, your relationships, your expertise — is rarely what takes the most time. What takes time is the setup. The blank page. The reformatting. The first draft. The summary you have to write before the meeting. The three-paragraph email you spend 40 minutes on because you’re not sure how to start.

That’s the friction. And AI is exceptionally good at removing it.

The professionals who save the most time aren’t using AI to replace their thinking. They’re using it to skip the 70% of any task that’s purely mechanical, so they can spend more time on the 30% that actually requires them.

Once you see work through that lens, the hours start adding up very quickly.

The Four Workflows That Move the Needle

These aren’t hypothetical use cases. These are what I see my students implement — and report back on — week after week.

Reports and Proposals: From 4 Hours to 45 Minutes

Most professionals I work with spend somewhere between 3 and 5 hours writing a substantive report or proposal. The blank page is the enemy. They know what they want to say, but translating that into a structured document takes forever.

The fix is counterintuitive: don’t ask AI to write the report. Ask it to help you structure your thinking first.

I have my students start by doing a brain dump — bullet points, rough notes, whatever’s in their head. Then they feed that into ChatGPT or Claude with a prompt like: “Here are my raw notes. Help me identify the key argument, the supporting points, and the best structure for a proposal to [specific audience].” The AI organises their thinking back at them.

From there, they draft section by section, using AI as a co-writer rather than a ghostwriter. The result is still their work, their expertise, their voice — but the scaffolding went up in minutes instead of hours.

Try this right now. Think of a report, proposal, or document you’ve been putting off. Do a brain dump — just bullet points, whatever’s in your head — and paste it into ChatGPT with this prompt:

Here are my raw notes on [topic]. They are unorganised — just capture everything I know.

[Paste your bullet points here]

I need to turn this into a [report / proposal / executive summary] for [audience — e.g., "my CEO," "a potential enterprise client," "the board"].

Please:
1. Identify the single most important argument or recommendation I'm making
2. Suggest a logical structure (section headings only)
3. Flag any gaps — things my audience will probably want to know that I haven't covered

Do not write the draft yet. Just help me think through the structure.

This takes about five minutes and saves the two hours of staring at a blank page.

Meeting Prep: From 2 Hours to 30 Minutes

Preparing for a high-stakes meeting used to mean two hours of research, slide review, and mental rehearsal. My students now do it in 30 minutes.

The workflow: drop the meeting agenda, any relevant context, and the names and roles of key attendees into AI, then ask it to anticipate the top five questions or objections each stakeholder is likely to raise. Ask it to draft talking points for each. Ask it to write the pre-read memo.

You still review everything. You still adjust for what you know about the room. But instead of starting from nothing, you’re editing from a solid draft. That’s the difference.

Email Volume: From 1 Hour Per Day to 15 Minutes

Email is the one that surprises people most, because it feels fast — until you add it up. One hour a day across 48 working weeks is 240 hours a year. That’s six full working weeks, gone.

The approach I teach: AI drafts, you edit. You give it the context, the recipient, the outcome you want, and it produces a first draft. You spend 90 seconds refining it instead of 10 minutes staring at a blank compose window.

The key detail here is building a personal style prompt. Your emails sound like you — a specific you, with a particular tone and level of formality. If you write that into a reusable prompt (“My communication style is direct, warm, and professional. I avoid jargon. I usually open with context before making an ask.”), the AI outputs land much closer to what you’d write yourself. My students build this in Week 1 and use it for the rest of the bootcamp.

Content and Communications: From a Full Day to 2–3 Hours

Many of my students are responsible for a monthly newsletter, LinkedIn posts, or internal communications. Before the bootcamp, producing a month’s worth of content would eat an entire day — sometimes more.

The shift isn’t just about using AI to write faster. It’s about using AI to structure the narrative before you start writing. What’s the core idea? What does the reader already believe? What do you want them to think, feel, or do after reading? Once you’ve answered those questions with AI’s help, the actual writing is much faster — and the output is better, because you started with a clear architecture.

Two to three hours for a newsletter plus a week of social posts is now normal for my graduates. The content is stronger too, because the thinking is cleaner.

The Mistake Most People Make

When professionals first start using AI, they try to automate too much. They want to hand the whole task over — paste in a brief and get a finished deliverable back.

Sometimes that works. But the people who save the most time over the long run are not the ones who automate the most. They’re the ones who stay in control of the 30% that requires judgment, and use AI to handle the 70% that doesn’t.

That distinction matters. A report that sounds like AI wrote it — because AI did write it, without any real human guidance — erodes trust with your stakeholders. A report that reflects your thinking, structured and drafted with AI’s help, still sounds like you. That’s the version worth sending.

Why This Is Especially Relevant in Hong Kong

I want to be specific about something, because I think it often goes unsaid.

English-language AI tools are disproportionately powerful in Hong Kong’s corporate context. The training data behind models like GPT-4 and Claude is overwhelmingly English, which means the quality of output for English-language tasks — drafting, editing, summarising, structuring — is genuinely exceptional.

Most of my students work in global companies or deal with international stakeholders. Their reports go to London or New York. Their proposals are in English. Their executive communications are in English. That puts them in a particularly strong position to capture the benefit of these tools, right now, before adoption becomes universal.

The professionals who build these habits today will be operating at a different level in two years. Not because AI will do their jobs for them — but because they’ll have eliminated a category of low-value work that their peers are still grinding through.

How I Measure Whether This Actually Works

I don’t measure success in my bootcamp by whether students understood the material. I measure it by whether they visibly changed how they work by Monday.

Every week, students come back and report what they tried. What worked. What didn’t. Where they got stuck. The time savings I’ve described above — 60 hours down to 15, reports from 4 hours to 45 minutes, email from 1 hour to 15 minutes — these are not projections or estimates. They’re numbers that came back from actual participants, tracked against how they were working before.

That’s the benchmark I hold myself to. If a student finishes the four weeks and doesn’t have at least two or three workflows that have genuinely changed how they operate, I haven’t done my job.

If You Want These Workflows for Your Team

The tools exist. The techniques are learnable. What most people lack is the structure to actually implement them — and accountability to do it under deadline.

That’s what the bootcamp provides. Four weeks, practical workflows every week, and a direct line to ask questions as things come up.

If you want these workflows for your team, I run a 4-week AI In Action Bootcamp in Hong Kong. Teams of 5–10 welcome. You can find out more at AI In Action Bootcamp.

The hours are there to be recovered. The question is whether you want to be the one recovering them — or watching someone else do it first.