## 1

1. That Familiar Thursday Night

Many of us have been there: you have a 15-minute product presentation on Friday morning, but only find time to prepare on Thursday night. You open a blank PPT, and just deciding "what goes on the first slide" stumps you for ten minutes. Writing the outline, finding templates, aligning layouts, drafting speaker notes, timing your rehearsal — the whole process easily takes four or five hours, and most of it is spent on "moving things around": transferring content already in your head into text boxes, then adjusting font sizes and line spacing.

For indie developers and small teams, this problem is even sharper. You probably come from a technical background and build things quickly, but "packaging" and "presentation" aren't your strong suits. A logically rigorous but visually messy PPT loses persuasive power in front of clients and investors. Hiring a designer? The budget doesn't allow it, and there's no time anyway.

My old workflow looked roughly like this:

• Write the outline: 30–45 minutes, with repeated edits
• Build the PPT: 2–3 hours, mostly burned on layout and alignment
• Write speaker notes: 40–60 minutes
• Rehearse and adjust: 30+ minutes

Total: about 4–5 hours — and the resulting PPT often still ended up "text-heavy, flat in structure, with no clear emphasis."

2. What AI Can Do on This Assembly Line

Break PPT-making into four stages, and you'll find AI can substantially take over the work at every stage:

1. Outline generation and structural optimization. Just feed your raw material (a product description, a requirements doc, or even a few transcribed voice notes) to AI and have it output a structured outline: each slide's title, core argument, supporting information, and suggested presentation format (chart / comparison / quote). The biggest value AI brings at this stage isn't "writing fast" — it's that it reviews the structure from the audience's perspective: where background context is missing, where evidence is too thin, where slides should be merged. This step shrinks from 45 minutes to about 5 minutes of conversational iteration.

2. Content filling and copy polishing. Once the outline is confirmed, have AI generate 2–3 versions of copy per slide for you to choose from. PPT copy differs from articles: it needs to be short, have hooks, and support your spoken delivery. AI can quickly produce multiple versions for comparison, whereas when you write it yourself, the first draft is usually the final draft — no options at all.

3. Generating speaker notes in sync with the slides. This is a stage many people overlook: speaker notes shouldn't be written from scratch after the PPT is done — they should be generated alongside each slide's content. Have AI output in the format "each slide + corresponding script + estimated duration," and your rehearsal pacing will be much better. In my actual tests with a 15-minute presentation, AI's first-draft script was within 1 minute of the target duration — just minor manual tweaks needed.

4. Filling the gaps. Have AI play the role of a "picky audience member" and ask pointed questions about your PPT. This step exposes weaknesses before the actual presentation — my old workflow had no such step at all. No time for it.

3. My New Workflow: The 40-Minute Version

The current workflow:

1. 0–5 minutes: Feed the raw material to AI, generate a draft outline, then manually adjust structure and order
2. 5–10 minutes: Generate copy slide by slide, pick versions, tweak details
3. 10–25 minutes: Paste into a PPT tool for layout (or use an AI-powered PPT tool to produce a draft directly and then edit it)
4. 25–35 minutes: Generate the accompanying speaker notes, read through and fine-tune
5. 35–40 minutes: Have AI run an "audience stress test" and add a Q&A contingency slide

Total: about 40 minutes — roughly 85% time savings compared to the old workflow. The more important change is quality: clearer structure, speaker notes strictly aligned with slides, plus the "rehearsal stress test" step that was previously impossible.

The time saved translates directly into money for a small team: someone who makes one PPT per week saves over a hundred hours a year — the equivalent of two or three full work weeks.

4. A Reusable Prompt Template

The template below is the version I've iterated on many times. You can copy it directly and replace the content in brackets:

你是一位资深演示设计顾问。请基于以下材料帮我完成一份PPT的完整策划。

【材料】
[粘贴你的原始材料：产品说明/项目文档/会议记录等]

【基本信息】
- 演讲场景：[客户汇报/投资人路演/团队复盘/技术分享]
- 听众画像：[他们对主题的了解程度、最关心什么]
- 时长：[X分钟]，预计页数：[X页]
- 我的角色：[产品负责人/开发者/...]

请按以下要求输出：

1. 大纲：每页包含「页码 | 页面标题 | 核心信息(一句话) | 呈现形式建议(要点列表/图表/对比/引用) | 该页目的(为什么听众需要它)」
2. 每页文案：标题控制在10字内，正文要点每条不超过20字
3. 配套讲稿：每页一段口语化讲稿，标注预计时长，总时长匹配我的要求
4. 听众预判：列出听众最可能提出的3个尖锐问题及应答要点
5. 结构自检：指出这份大纲中最薄弱的一页，并给出改进建议

先只输出大纲，等我确认后再输出第2-5项。

The final "step-by-step confirmation" design is critical: if you ask AI to output everything at once, you'll get a pile of content that needs major rework. Lock in the outline first, and the quality of everything that follows goes up a notch.

5. A Few Pitfall Warnings

• Don't use AI's layout drafts as-is. Generative PPT tools produce a heavy template feel. Manually adjusting layout and visual emphasis is a step you can't skip, but it only takes about 10 minutes.
• Always read the speaker notes aloud yourself. AI's duration estimates are based on average speaking speed; only a rehearsal can calibrate for your actual pace and pausing habits.
• Manually verify all data. When filling in examples and data, AI tends to "make things up convincingly." Every number must be traced back to its source.
• On tool costs: If you do this by calling LLMs via API, prices vary significantly between models — refer to the official pricing pages. For a task like making a PPT, usage is low; a full planning session typically consumes very few tokens. It's a "pays for itself on the first use" scenario.

6. Final Thoughts

Making a PPT is essentially the process of "reorganizing existing information into a persuasive structure." You already have the information; what's missing is structural skills and the time for lots of repetitive labor — which is exactly where AI excels at filling in. Compressing 4–5 hours down to 40 minutes isn't about making more PPTs; it's about spending the saved time on what genuinely needs a human: figuring out who you're trying to persuade, and why.

If you haven't yet tried building your own content workflow with AI APIs, start with a gateway service that supports multiple models — switch models on demand, pay as you go, with very low trial-and-error cost. Here's a recommended registration link: https://api.thistoken.ai/register. Once registered, you can call mainstream LLMs, run the prompt template above directly, and experience the efficiency of "delivering in 40 minutes."

---

Ready to try it yourself? Sign up at https://api.thistoken.ai/register to get your API key and start building.