Work Time Is Made of "Writing" and "Reading"
Think back over your working hours. Writing email, putting together minutes, drafting reports, reading research materials, checking proposals—most work consists of time spent writing text and time spent reading text. Those two are exactly what generative AI is good at. This chapter starts by giving you the big picture of how to use AI to speed up both "writing" and "reading."
This article works on its own, but having a map of "which tool fits what" up front lets you read the rest of the chapter (minutes tools, translation tools, proofreading tools, agent use) without getting lost.
4 Patterns to Speed Up "Writing" with AI
1. Create a Draft
Starting from a blank page is the most time-consuming part. Having AI produce the first draft of "a progress report to your manager," "the skeleton of a proposal to a client," or "an internal announcement" removes the time of thinking from zero. The basis is to receive AI's text as a rough draft and fix it yourself, not send it as is.
2. Change the Tone
Even the same content needs a different delivery depending on the recipient. Paste your text and ask "make this formal for executives," and it rephrases into "polite keigo," "a casual internal tone," or "English business writing" in seconds.
3. Proofread and Revise
Beyond typos and notation inconsistency, you can have it point out "whether the point comes across" and "whether the logic jumps." A good trick is to ask "name 3 spots that are unclear from the reader's perspective."
4. Translate
It handles Japanese–English both ways and other major languages. Because translations reflect context, you can polish into "English that reads naturally as a business email" rather than word substitution.
4 Patterns to Speed Up "Reading" with AI
1. Summarize
Shrink long articles, PDFs, and minutes to 3 lines to half a page. Specifying an output template like "conclusion first, 3 grounds, 1 supplement line" yields consistent quality every time.
2. Extract
You can pull out only the information you need, like "only the remaining tasks as bullets" or "only the numbers that appeared, as a table." You get just the part you want without reading the whole thing.
3. Read While Translating
You can read through English papers and articles while translating. Having it explain jargon at the same time speeds comprehension even in unfamiliar fields.
4. Read Charts and Tables
Paste a screenshot or a PDF chart and ask "give me 3 things this table shows." Because it understands images too, it works on chart-heavy materials.
Which AI for What (as of May 2026)
The main chat AIs are ChatGPT (OpenAI, the GPT-5.5 line in 2026) and Claude (Anthropic, the Opus 4.7 line). As of 2026 both can handle roughly one million tokens of text at once, making it realistic to load long materials. Everyday tasks—drafting, summarizing, translating—work with either. On top of that, adding a dedicated tool for a given use raises accuracy and speed.
| What you want | Recommended combination | Reason |
|---|---|---|
| Email/report draft | ChatGPT or Claude | Writes anything from casual to formal |
| Long-text/PDF summary | Claude (or ChatGPT) | Large context keeps the whole picture while summarizing |
| Translation | DeepL rough → polish with AI | Splits translation accuracy and business-tone tuning |
| Meeting minutes | Otter / Notta / tl;dv → format with AI | A dedicated tool handles recording to transcription |
| English proofreading while writing | A resident tool like Grammarly | Flags issues automatically as you type |
Tool names and prices are covered in detail in each article of this chapter. Plans and prices change often, so always check the official site for the latest before subscribing.
The Strongest Use Is "AI Drafts, Human Finishes"
Demanding perfection from AI ironically makes checking take longer. The realistic split is "9/10 in 30 seconds, the remaining 1/10 by yourself in 5 minutes." AI is weak in the following areas; assume a human tidies these at the end.
- Company-specific culture and etiquette (in-house phrasing, the order of approvals)
- Sensitive internal circumstances (inter-department dynamics, hard-to-say background)
- Things requiring fact-checking (AI sometimes writes non-facts plausibly; verify numbers, proper nouns, and dates against the source)
Common Misunderstandings
- "AI can do everything if I leave it to it": drafting and summarizing are fast, but final judgment and fact-checking are the human's job.
- "Just sending 'summarize this' is enough": without specifying purpose, reader, and output format, you only get a mediocre short text. The next article covers templates in detail.
- "I can paste confidential materials as is": for internal-only or customer personal data, check the service's data-handling terms and anonymize if needed before passing it.
The Next Article
The next article, "Handling Long Documents: Report Summarization, Paper Reading, and Book Analysis," covers concrete steps and prompt examples for having AI read long materials. After that, the chapter proceeds to minutes automation, English business documents, dedicated tools (Otter / DeepL / Grammarly / Notion AI), and finally applications where an agent handles email replies and long-document creation.
