Personal AI workflow construction logic: first **identify high-frequency repetitive tasks** (daily/weekly information processing, writing, analysis tasks) → **assess AI replaceability** (pure text processing > requiring professional judgment > requiring external data > requiring interpersonal relationships) → **design AI intervention nodes** (AI drafts initial version, AI summarizes information, AI generates options → human review and decision) → **build a Prompt template library** (save validated effective Prompts as reusable templates).
## Major AI Tool Matrix
**ChatGPT (OpenAI)**: most widely used general LLM; GPT-4o excels at multimodal (image understanding) and code generation; suitable for daily writing, brainstorming, code debugging; Plus subscription (approximately $20/month) unlocks GPT-4o and plugins. **Claude (Anthropic)**: long context window (200K tokens), particularly strong at long document analysis, complex writing tasks, following complex instructions; Projects feature lets users manage long-term project context. **Gemini (Google)**: deeply integrated with Google services (Gmail, Drive, Docs); suitable for workflows needing smooth Google ecosystem integration; Gemini Advanced has strong reasoning capability. **Local models (Ollama + Llama/Mistral)**: run on local machines; suitable for scenarios requiring data privacy protection (not uploading internal files to third-party servers). **Specialized AI tools**: Cursor (AI code editor), Perplexity (AI search), Midjourney/Sora (image/video generation), Otter.ai (meeting notes) — each with niche scenario advantages.
## High-Value AI Workflow Examples
**Daily/weekly report automation**: collect week’s data (export from database/Airtable) → use Claude/GPT to summarize key metric changes → generate initial draft → 5-minute human review → send. **Competitive intelligence monitoring**: set Google Alerts (competitor company names) → collect daily news → AI summarizes key developments → outputs briefing. **Meeting preparation**: upload meeting-related documents → AI generates agenda summary, key questions, and background information cards → read AI briefing 5 minutes before meeting. **Content creation pipeline**: AI generates content outline → human filters and confirms → AI generates initial draft per outline → human review + fact-checking → publish.
See [Prompt Engineering in Practice](https://sunqi.org/prompt-engineering-guide-en/) and [Anthropic Claude Documentation](https://docs.anthropic.com/).
