Claude for Daily PM Work: Basic to Advanced
Use Claude every day for writing, research, analysis and prototypes, then connect it to Gmail, Calendar, Drive, Slack, GitHub and your own tools.
On this page
- What Claude is and where you can use it
- Level 1 · Your first hour with Claude
- Level 2 · Features working PMs use every week
- Level 3 · Connect Claude to your apps
- A daily routine for PMs
- Labs
- Prompt library for PMs
- Common mistakes
- Where to find official help and sources
- Exercises with model answers
- Key takeaways
- Check your understanding
- Recommended resources
Feature names move around
Claude ships new features often, and menus get renamed. If a button in a lab has moved, search the Claude Help Center for the feature name. Screens below are illustrative wireframes, not screenshots.
14.1 What Claude is and where you can use it
Claude is Anthropic's AI assistant. You talk to it in plain language, give it files, and it writes, analyses, researches, codes and builds small apps. The same account works in several places:
Web: claude.ai
- Every feature
- Best for long work sessions
- Artifacts, Projects, connectors
Desktop app (Mac, Windows)
- Same as web, plus local extensions
- Keyboard shortcut to open quickly
- Works with files on your laptop
Mobile app (iOS, Android)
- Voice mode for hands-free thinking
- Photo of a whiteboard → summary
- Good for quick capture
Claude Code
- Terminal, IDE, desktop and web
- Works inside code repositories
- Class 15
Inside other apps
- Slack, browser extension and others where your plan offers them
- Check Settings → Connectors and the Help Center
API (Claude Console)
- Build Claude into your own app or bot
- Pay per use
- Used in Classes 20 to 22
14.1.1 Plans
| Plan | Who it is for | What you get (checked 23 Oct 2026) |
|---|---|---|
| Free | Trying Claude | Chat, file upload, limited daily messages, some connectors |
| Pro | Individual PMs | More usage, Projects, more models, Claude Code included |
| Max | Heavy daily users | 5× or 20× Pro usage, priority access, early features |
| Team / Enterprise | Companies | Admin controls, shared Projects, single sign-on, data controls |
Chat and Claude Code draw from the same usage pool on Pro and Max, so a long coding session leaves less chat time that day. Check claude.com/pricing for current limits.
14.2 Level 1 · Your first hour with Claude
Level 1 · Beginner
Goal: get one genuinely useful result (a meeting summary, an email, a first PRD draft) and learn the four habits that make answers better.
14.2.1 The anatomy of a good prompt
Most weak answers come from missing context, not from a weak model. Use four parts:
14.2.2 Four habits for Level 1
- Give the source material. Attach the PDF, CSV or notes instead of asking from memory.
- Ask for a format. Tables, bullet lists, "max 150 words", "for a VP audience".
- Iterate in the same chat. "Shorter", "more specific", "add a risks section".
- Start a new chat for a new topic. Long mixed chats get slower and less focused.
14.2.3 Files Claude can read
PDFs, Word documents, slides, spreadsheets and CSVs, images and screenshots, and plain text or code files. Claude can run code to analyse data files: upload a CSV and ask for averages, a chart or a cohort table.
Data analysis prompt
"Here is orders.csv. 1) Describe the columns. 2) Weekly order count and average order value. 3) A line chart of both. 4) Three observations a PM should act on. Show the method, so I can check it."
14.3 Level 2 · Features working PMs use every week
Level 2 · Working PM
Goal: stop re-explaining your product in every chat, produce documents and prototypes you can share, and research with sources.
14.3.1 Projects: your product's memory
A Project is a workspace with custom instructions and knowledge files that every chat inside it can use. Make one Project per product or per big initiative.
14.3.2 Artifacts: documents, code and clickable prototypes
When Claude produces something substantial (a document, a diagram, a web page, a small app), it appears as an artifact in a panel beside the chat. You can edit it by asking for changes, see versions, copy or download it, and publish it to get a link you can share.
14.3.3 Research with sources
Turn on web search (and Research, where your plan offers it) for questions that need current information: competitors, regulations, market sizes. Research runs many searches and returns a report with citations.
Always open the sources
Click at least three citations and confirm the claim is really there. AI research can misread a source or rely on a weak one. Mark any number you could not verify as "unverified" in your document.
14.3.4 Styles, memory and skills
- Styles set how Claude writes (concise, formal, explanatory, or a custom style learned from your own writing samples).
- Memory and past-chat search (where enabled) let Claude recall earlier work. You can view, edit or turn memory off in Settings.
- Skills are packaged instructions (and sometimes scripts) that teach Claude a repeatable task, such as your company's PRD format or a spreadsheet report. Claude loads a skill when the task matches. You can use Anthropic's skills and create your own.
14.4 Level 3 · Connect Claude to your apps
Level 3 · Advanced
Goal: let Claude read your email, calendar, documents and Slack, act with your approval, and use your own tools through MCP.
14.4.1 Connectors
Connectors give Claude permission to use another app on your behalf. Go to Settings → Connectors, click Connect next to a service, sign in, and approve the permissions.
14.4.2 What PMs do with connectors
| Connector | Daily use | Example prompt |
|---|---|---|
| Gmail | Triage, find decisions, draft replies | "Find every email about the payments launch this month. List decisions and open questions, then draft a reply to the latest thread." |
| Google Calendar | Prep and planning | "What meetings do I have tomorrow? For each, what should I prepare, based on related Drive docs?" |
| Google Drive | Search and summarise docs | "Compare the Q3 and Q4 roadmap docs and list what moved." |
| Slack | Catch up and draft updates | "Summarise #checkout-squad since Monday: decisions, blockers, who is waiting on me." |
| GitHub | Read code and issues | "Read the open issues labelled 'checkout' and group them by theme." |
| Notion / Figma | Specs and designs | "Pull the requirements from the Split Bill PRD in Notion and check the Figma file covers each one." |
14.4.3 Custom connectors and MCP
MCP (Model Context Protocol) is an open standard that lets an AI assistant use tools from another system. A remote MCP server has a URL. Add it under Settings → Connectors → Add custom connector, and Claude gets that server's tools. In Class 20 you will build an n8n workflow with an MCP Server Trigger and add it to Claude, so you can say "log this bug to our tracker" and n8n does it.
14.4.4 The API and Claude Console
To put Claude inside your own app, bot or automation, you use the Claude API:
- Sign in at console.anthropic.com (separate billing from your claude.ai plan).
- Add a payment method and set a monthly spend limit.
- Create an API key and store it in a secret store (never in a chat or frontend code).
- Test prompts in the Workbench, then use the key in n8n, Make, Zapier, Lovable or code.
14.5 A daily routine for PMs
| Time | Task | Feature | Prompt starter |
|---|---|---|---|
| 09:00 | Inbox and calendar brief | Gmail + Calendar connectors | 'Brief me: urgent emails, today's meetings, what to prep' |
| 10:00 | Synthesise user interviews | Project + file upload | 'Top themes with quotes and counts; flag contradictions' |
| 11:30 | Draft or tighten a PRD section | Project + PRD skill or template | 'Critique this scope against our strategy file' |
| 13:00 | Analyse a metric drop | File upload + analysis | 'Explain the conversion drop by segment; show the method' |
| 15:00 | Prototype an idea for design review | Artifact | 'Clickable mobile prototype, sample data, three states' |
| 16:30 | Competitor or market question | Research with citations | 'How do the top 3 rivals handle refunds? Cite sources' |
| 17:30 | Weekly or daily update | Slack connector | 'Draft my update from #squad and my notes, 120 words' |
14.6 Labs
Lab 1 · Build your product Project (20 minutes)
- Projects → New project → name it after your product.
- Write instructions: product in one line, users, goals this quarter, writing rules, "flag assumptions".
- Upload three files: a strategy or vision doc, personas, and one metrics CSV (use
datasets/orders.csvfrom this handbook if you have nothing). - Ask: "Using only the knowledge files, what are the three biggest risks to this quarter's goal?"
- Check the answer cites the files. If it invents facts, add to the instructions: "Answer only from the knowledge files; say 'not in files' otherwise."
Lab 2 · Interview synthesis (30 minutes)
- Inside your Project, attach five interview notes (real or written by you).
- Prompt: "Affinity-map these interviews: themes, number of users per theme, one quote each, and the job-to-be-done behind each theme."
- Follow up: "Which theme is most likely to change our roadmap? What would we need to believe?"
- Save the result as an artifact document and publish it for your team.
Lab 3 · A clickable prototype you can share (30 minutes)
- Prompt: "Build a mobile-first clickable prototype for [feature]: three screens, sample data only, a 'Prototype · not real' label."
- Ask for two changes, one at a time.
- Ask for an edge case: "Add the empty state and an error state."
- Click Publish, open the link in a private window, and check it works on your phone.
Lab 4 · Morning brief from Gmail and Calendar (15 minutes)
- Settings → Connectors → connect Gmail and Google Calendar (use a practice account first).
- In a new chat, turn both on in the tools menu.
- Prompt: "Brief me for today: meetings with what to prep, emails that need a reply from me, and anything with a deadline this week."
- Ask Claude to draft two replies. Open Gmail and check the drafts are there. Claude drafts; you review and send.
Lab 5 · Weekly Slack update (15 minutes)
- Connect Slack (practice workspace).
- Prompt: "Read #practice-squad from the last 7 days. Draft a weekly update: shipped, in progress, blocked, decisions needed. 120 words."
- Edit the draft yourself before posting. You own what you send.
Lab 6 · Give Claude a custom tool (advanced, after Class 20)
- Build the n8n "log a bug to Google Sheets" workflow with an MCP Server Trigger (Class 20, Lab 5).
- Copy its production URL.
- Claude → Settings → Connectors → Add custom connector → paste the URL.
- In a chat: "Log a bug: checkout button overlaps on small screens, high priority." Approve the tool call, then check the sheet.
14.7 Prompt library for PMs
| Job | Prompt |
|---|---|
| PRD critique | "Act as a skeptical engineering lead. List the 10 questions you'd ask about this PRD, then rewrite the scope section to answer them." |
| Prioritisation | "Score these 12 ideas with RICE. Show your assumptions for reach and effort, and mark which ones are guesses." |
| Metrics | "Build a KPI tree for 'weekly active ordering users'. Include input metrics I can move this quarter." |
| Stakeholders | "Rewrite this update for (a) the CEO in 3 lines, (b) engineers with the technical detail." |
| Interview prep | "Ask me one PM interview question at a time on product sense. After my answer, score it and show a stronger answer." |
| Release notes | "Turn these merged pull request titles into customer-friendly release notes, grouped by benefit." |
14.8 Common mistakes
- No context: asking "write a PRD" with no product, users or goal. Use a Project.
- Trusting numbers and quotes without checking: always verify against the source.
- Pasting confidential data into a personal plan: follow your company's AI policy; use Team or Enterprise for work data.
- One endless chat: start a new chat per task; keep long-lived context in the Project.
- Letting Claude send on your behalf too early: keep approvals on for email and Slack actions.
14.9 Where to find official help and sources
- Docs: docs.claude.com (API, prompting, tools, MCP, Claude Code).
- Help Center: support.claude.com (features, plans, connectors, troubleshooting).
- Claude Academy (free; formerly Anthropic Academy at anthropic.com/learn, which reportedly now redirects there): courses including Claude 101, AI Fluency, Building with the Claude API, Introduction to MCP and Claude Code in Action at academy.claude.com.
- Prompt engineering guide: in the docs, under "Prompt engineering".
- Cookbook: github.com/anthropics/anthropic-cookbook for API examples.
- Anthropic on YouTube: official talks and feature walkthroughs.
14.10 Exercises with model answers
Exercise 1 · Fix the prompt
"Write a PRD for a loyalty program." Rewrite it using the four-part prompt.
Model answer
"I'm the PM for a grocery delivery app in Indian metros; 60% of orders come from repeat users (context). Draft a one-page PRD for a points-based loyalty program aimed at increasing 90-day retention (task). Use our PRD template and strategy file in this Project (input). Sections: problem, goals and metrics, users, scope in and out, risks, open questions; max 600 words; flag every assumption (format)."
Exercise 2 · Which feature?
You need (a) a shareable clickable concept for tomorrow's design review, (b) Claude to remember your product's personas in every chat, (c) a summary of last week's Slack decisions. Which Claude feature solves each?
Model answer
(a) An artifact, published as a link. (b) A Project with the personas in knowledge files and the instructions. (c) The Slack connector, enabled in that chat.
14.11 Key takeaways
- Good results come from context, a clear task, source material and a format.
- Projects hold context; artifacts turn answers into shareable documents and prototypes.
- Connectors let Claude work with Gmail, Calendar, Drive, Slack, GitHub and more, with your approval.
- Custom connectors (MCP) and the API extend Claude to your own tools and apps.
14.12 Check your understanding
Check your understanding
You keep pasting the same personas and strategy into every chat. What should you use?
- A longer prompt
- A Project with instructions and knowledge files
- The API
- A new account
With the Gmail connector, what does Claude do with a reply you ask for?
- Sends it immediately
- Creates a draft for you to review and send
- Posts it to Slack
- Nothing; Gmail is read-only
How do you give Claude a tool from your own n8n workflow?
- Paste the n8n password into a chat
- Add the workflow's MCP server URL as a custom connector
- Upload the workflow as a PDF
- It is not possible
Where should your Claude API key live when you use it in an automation?
- In the chat
- In the automation tool's credential store
- In a public GitHub repository
- In the artifact code
14.13 Recommended resources
- Claude Help Center · Claude docs · Claude Academy
- Connectors: getting started
- Claude Console for API keys and spend limits
- Anthropic cookbook on GitHub