When you feel the text generated by AI is 'just not quite right.' Prompt design techniques for meeting minutes, reports, and proposals

Introduction | 'I asked the AI, but it only gave me unusable text'

'I tried it once, but it only gave me text that felt a bit off...'

Have you ever felt this way when using AI?

When I entered 'create meeting minutes' into ChatGPT, it gave me irrelevant text.
When I asked it to 'write a report,' it was too generic to be useful.
I just stopped trying.

For 90% of people who struggle with AI, it is not a problem with the AI's capabilities.

It is a problem with 'how to write prompts (instructions for AI).'

AI is like a talented subordinate.
If you give vague instructions, you get vague results.
If you give specific instructions, you get surprisingly useful results.

In this note, I will share design techniques for 'usable prompts' using three scenarios: meeting minutes, reports, and proposals.

Why the output becomes 'unusable'

If instructions to the AI are vague, the AI will return a 'generic answer'.

'Create meeting minutes'
→ Since it doesn't know what kind of meeting it is, it outputs a generic meeting minutes format

'Write a report'
→ Since it doesn't know who it is for or what it is about, it outputs vague text

'Think of a structure for a proposal'
→ Since it doesn't know what the proposal is for, it outputs a textbook-style structure

This is not because the AI is 'useless,' but simply because'information about the situation, the target audience, and the objective' is missing.

Prompt design is what fills this gap.

Learning the '4 elements' of a prompt

Usable prompts all share four common elements.

① Role | What position do you want the AI to answer from?
Example: 'You are a meeting facilitator for a small business'

② Situation | Background, context, and constraints
Example: 'Sales meeting with 5 employees. 60 minutes. I will provide the meeting notes'

③ Instruction | What do you want it to do and how (specifically)?
'Organize by separating decisions, actions, person in charge, and deadlines'

④ Format | Output shape, length, and tone
'Bullet points. Length that fits on one A4 page. Polite tone for internal sharing'

Just by being conscious of these four elements, the quality of the output will change dramatically.

Now, let's try it in an actual business scenario.

Practice ① Prompt design for meeting minutes

Before (Common vague prompt)

以下のメモを議事録にしてください。

・新商品の件、来月中に決める
・田中さんが担当
・予算どうするか要検討
・次回は再来週

→ Output generated: A formal meeting minute framework is created, but the information is thin, and you end up having to rewrite it yourself.

After (Prompt incorporating the 4 elements)

あなたは会議の議事録作成の専門家です。

【会議の状況】
- 会議名：新商品開発会議
- 日時：2026年5月15日 14:00〜15:00
- 参加者：田中（営業部長）、鈴木（開発担当）、山本（経営企画）
- 目的：新商品ラインナップの最終決定

【私のメモ（箇条書き）】
・新商品A案とB案を比較→A案で進める方向に決定
・来月末（6月30日）までにサンプル完成が条件
・田中さんがクライアントへの事前説明を担当
・予算は開発費50万円の範囲内で収める→鈴木さんが確認して来週水曜までに報告
・次回会議：5月28日 同メンバーで

以下の形式で議事録を作成してください。
1. 決定事項（箇条書き）
2. 検討継続事項（箇条書き）
3. 次のアクション（担当者・期限つきの表形式）
4. 次回会議の日程
社内共有用・丁寧な文体で。A4で1枚以内に収めてください。

→ Output generated: Meeting minutes ready to send as-is. Revisions are limited to checking proper nouns.

Practical points for meeting minute prompts

You don't need 'perfect notes'.

During the meeting, just jotting down fragmented keywords and numbers is fine.
If you decide to 'leave the formatting to AI,' the stress of taking notes during meetings will be drastically reduced.

Information you want to add later can simply be appended to the 'My Notes' section of the prompt.

Practice ②: Monthly report prompt design

Before (Common, vague prompt)

先月の営業報告書を書いてください。
売上は1200万円でした。

→ Output generated: A thin text like, 'Last month's sales were 12 million yen. I will continue to work hard.'

After (Prompt incorporating the 4 elements)

あなたは中小企業の営業部門の月次報告書の作成を支援するアシスタントです。

【報告の状況】
- 報告先：社長（経営者）への月次報告書
- 対象月：2026年4月
- 報告者：営業課長

【今月の実績データ】
- 売上：1,200万円（目標比95%・前月比+8%）
- 新規顧客：3社（目標5社）
- 既存顧客の解約：1社（建設業・理由：予算削減）
- 商談件数：22件（うち成約8件、継続協議9件、失注5件）
- 主なトピック：A社との大口契約交渉が最終局面、来月成約見込み

【今月の課題】
- 新規開拓の件数が目標を下回った（営業リソースが既存フォローに割かれた）
- GW期間の商談減少が想定より影響した

【来月の予定】
- A社との契約締結（予定売上250万円）
- 新規開拓強化のため外部セミナー参加（2名）
- 月間目標：売上1,350万円

以下の構成で報告書を作成してください。
1. 今月の実績サマリー（数値込み）
2. 好調だった点と要因
3. 課題と対策
4. 来月の計画と期待値
簡潔・具体的に。経営者が5分で読める分量（A4で1〜1.5枚）。

→ Output generated: A draft report with organized numbers and context that can be submitted as-is.

Practical points for monthly report prompts

If you use the same structure every month, I recommend saving it as a template.

The only things that change each month are 'Performance Data,' 'This Month's Issues,' and 'Next Month's Plans.'
If you only rewrite those parts, a draft report can be completed in 5 minutes every month.

Practice ③: Prompt design for creating proposal and project plan structures

'I have to create a project plan, but I don't know how to structure it.'
'I want to write a proposal, but I don't know where to start.'

In these cases, starting by having the AI 'think of the structure' makes it easier to get moving.

Before (Common, vague prompt)

社内研修の企画書を書いてください。

→ Output generated: A generic training proposal format. It does not reflect your company's situation at all.

After (Prompt designed with 4 elements in mind)

あなたは中小企業の人事担当者をサポートする企画立案アシスタントです。

【企画の状況】
- 対象：社員20名の製造業（北陸・富山）
- 企画内容：管理職（課長・係長クラス5名）向けの「評価面談スキル向上研修」
- 背景：今期から人事評価制度を刷新。評価面談のやり方が統一できていないという課題がある
- 予算感：外部講師を入れずに内製でやりたい。外部研修費用の補助金活用も検討中
- 実施時期：7月中（土日は避けたい）
- 決裁者：社長（60代・現場主義・コストに敏感）

以下の形式で企画書の構成と各セクションの骨子を作成してください。
1. 企画の背景・目的（社長に刺さる課題設定）
2. 研修の概要（対象・日程・形式・期待成果）
3. プログラム案（半日×2回 or 1日×1回の2パターン）
4. 予算概算（内製コストの試算）
5. 補助金活用の可能性（簡単に）
6. 期待される効果（定量・定性）
社長への報告書として提出するイメージで。承認を得やすい構成にしてください。

→ Resulting output: The framework of the proposal is complete. All that's left is to verify the numbers and flesh it out.

Practical points for proposal prompts

For proposals, it is important to solidify the 'structure' first.

If you try to make the AI create a finished product in one go, the content will be thin.
The secret to creating high-quality proposals is a two-step process: first, have it create the 'structure and outline,' and then use that as a draft to flesh it out yourself.

If you include the 'characteristics of the decision-maker' in the prompt, the AI will propose a structure that is conscious of 'phrasing that resonates with this person'.

5 common failure patterns

Failure 1: Expecting 'something good on the first try'

AI is a conversational tool. It is natural that the first output is not 100% perfect.

By repeating exchanges like
'Make this part more specific'
'Change this expression'
the accuracy will improve.Don't seek perfection in one go;use it with the feeling of 'having a rally'
to get results.

Failure 2: Thinking 'the AI will think for me' without providing information

AI is a creature that 'does not know about your company'.

Industry, scale, situation, counterpart, purpose
If you don't provide these, it will only return generic answers.
Remember that 'the more information you provide, the higher the accuracy'.

Failure 3: Using the output as-is

The text produced by AI is a 'draft'.

Checking proper nouns, accuracy of numbers, and adjusting to your company's tone.
These three things must always be checked by a human.
Don't forget the stance that 'the final responsibility for AI output lies with the human'.

Failure 4: Trying to write a prompt that is too long

If you think 'I must write the perfect prompt,' you will get exhausted creating prompts.

At first, it is enough to just provide 5 to 8 situations in bullet points.
If it's not enough, you can just ask for more.

Failure 5: Giving up if it doesn't go well once

'I tried it, but it wasn't quite right' is a path everyone goes through.

If it doesn't go well, it's just because the prompt was insufficient.
'Adding information',
'asking in a different way',
'changing the role'
will almost always improve accuracy.

AI is a tool that can only be mastered by those who do not give up.

In conclusion

I have noticed something while supporting AI utilization.

People who become good at using AI are 'people who are good at giving instructions'.

People who can clearly communicate 'what they want done and how' to others (whether subordinates or business partners) naturally become good at prompting AI as well.

Conversely, people whose instructions to AI are vague usually adopt a 'hope they just sort of understand' style in their daily communication as well.

AI is a mirror of your 'instructional ability'.

Refining your prompts also leads to refining your overall work communication;
it has that aspect to it.

One thing you can do starting today

When your next meeting ends, please try copying and using this text.

あなたはビジネス文書の専門家です。
以下の会議メモをもとに、社内共有用の議事録を作成してください。

【会議の基本情報】
・会議名：○○
・日時：○月○日 ○時〜○時
・参加者：○○、○○、○○

【メモ（箇条書き）】
・（ここに会議中のメモを貼り付ける）

出力形式：
1. 決定事項
2. 継続検討事項
3. 次のアクション（担当者・期限つき）
4. 次回会議日程
丁寧な文体で、A4で1枚以内に収めてください。

Try this once. It should change your perspective to 'AI might actually be useful'.

📌 For those who want to learn more deeply about AI utilization

In this note, I have conveyed the basics of prompt design.

I also explain the 'mindset of leaving meeting minutes to AI' in detail in a series of articles.

▶ 'Leaving meeting minutes, reports, and emails to AI—How I turned 30 minutes of work into 5 minutes'
(AI x Business Improvement Series Vol. 3)

Also, AI utilization in recruitment, human resources, and management tasks is currently being distributed in a separate series.

Detailed prompt collections and templates are scheduled to be introduced in future paid content.

📞 I will think about AI utilization implementation for your organization with you

"We want to expand AI usage in our company, but we don't know where to start."
"We want to teach our employees how to use prompts."
"We want to promote operational efficiency through AI and organizational design."

If you feel this way, please feel free to consult with us.
We offer a free 45-minute initial consultation.
We are available online and can also visit locations within the three Hokuriku prefectures.

▶ Click here to apply for a free consultation