AI Tools Landscape and Your Starter Stack
Which AI tool to use for which job, what each one costs, what you need before you start, and how the AI Tools Lab classes fit together.
On this page
- How to use the AI Tools Lab
- The AI tools map
- Which tool for which job
- What it costs: free versus paid
- What you need before you start
- Accounts, API keys and safety
- A starter stack for three budgets
- How the classes connect: one example project
- Exercises with model answers
- Key takeaways
- Check your understanding
- Recommended resources
Prices and features change every month
Every price, free limit and feature name in the AI Tools Lab was checked on 23 October 2026 against official pages or, where noted, several independent reviews. AI tools change quickly, so check the official pricing page before you pay or teach from these numbers. Screens in this lab are illustrative wireframes that show the layout and the order of clicks. They are not real screenshots, and button names may differ slightly in the live product.
13.1 How to use the AI Tools Lab
The AI Tools Lab is a set of hands-on classes. Each class follows the same pattern, so you can practise straight from the PDF:
- What the tool is and when to use it, in plain words.
- What you need: account, plan, cost, other tools.
- Level 1 (basic): your first result in under an hour.
- Level 2 (intermediate): the features working PMs use every week.
- Level 3 (advanced): connecting the tool to other apps, automating, and shipping.
- Labs: numbered steps with a picture of each screen or workflow.
- Mistakes, exercises with model answers, a quiz and resources.
13.2 The AI tools map
AI tools for PMs fall into eight groups. Most real projects combine two or three groups, for example Claude to write the spec, Lovable to build the app and n8n to connect it to Gmail and Slack.
1 · AI assistants
- Claude, ChatGPT, Gemini, Microsoft Copilot
- Think, write, analyse files, research
- Class 14
2 · Coding agents
- Claude Code, Cursor, GitHub Copilot, Codex
- Edit real code and repositories
- Class 15
3 · App builders
- Lovable, Replit, Bolt, v0, Figma Make
- Prompt to working web app
- Classes 16 and 17
4 · Design
- Figma, FigJam, Figma Make
- Screens, prototypes, design systems
- Class 18
5 · Docs, PRDs and planning
- Notion, ChatPRD, Confluence, Linear, Jira
- Specs, wikis, roadmaps
- Class 19
6 · Automation
- Zapier, Make, n8n
- Connect Gmail, Slack, Sheets, CRMs
- Class 20
7 · Chatbots
- WhatsApp Business Platform, Botpress, Voiceflow, n8n
- Answer customers in chat
- Class 21
8 · Voice agents
- Vapi, Retell AI, ElevenLabs Agents, Twilio
- Answer and make phone calls
- Class 22
13.3 Which tool for which job
| If you want to… | Start with | Free to try? | Upgrade when |
|---|---|---|---|
| Write a PRD, summarise interviews, analyse a CSV | Claude (Projects + file upload) | Yes | You hit message limits or need Claude Code |
| Change code in a real repository | Claude Code | Needs Pro or API | Daily use: Max or a team plan |
| Build a clickable app with login and a database | Lovable | Yes (5 credits a day) | You need private projects or a custom domain |
| Build and host a full-stack app in the browser | Replit Agent | Yes (limited) | You publish more than one app |
| Design screens and a click-through prototype | Figma | Yes (Starter) | You need Dev Mode or more files |
| Keep specs, roadmaps and research in one place | Notion | Yes | You need full Notion AI (Business plan) |
| Send a Slack alert when an email arrives | Zapier | Yes (100 tasks a month) | You need multi-step Zaps |
| Build complex automations cheaply | n8n (self-hosted) | Yes (free software) | You want managed cloud hosting |
| Answer customers on WhatsApp | A WhatsApp BSP + n8n or Botpress | Trial credits only | Real customer traffic (per-message fees) |
| Answer phone calls with AI | Vapi or Retell AI | Trial credits | Any real calls (per-minute fees) |
13.4 What it costs: free versus paid
The rule for this lab
You can complete every Level 1 lab for free. Most Level 2 labs are free too. Paid tools appear only where real messages, phone calls, or production hosting are involved, and each class says so clearly.
| Tool | Free tier | First paid tier | What you pay for |
|---|---|---|---|
| Claude | Free plan with limited messages | Pro, about $20 a month | More usage, Claude Code, more models; Max for heavy use |
| Claude API | No free tier for production | Pay per token | Building your own apps and bots on Claude |
| Lovable | 5 credits a day (max 30 a month), public projects | Pro, about $25 a month (100 credits) | Private projects, custom domain, more credits |
| Replit | Starter: daily Agent credits, 1 published app | Core, about $20 to $25 a month | More Agent work, more apps, collaborators |
| Figma | Starter: limited files and AI credits | Professional, about $16 a month per full seat | Dev Mode, MCP server, more Figma Make use |
| Notion | Free plan, limited AI trial | Business, about $20 per member a month | Full Notion AI, agents, meeting notes |
| Zapier | 100 tasks a month, two-step Zaps | Professional | Multi-step Zaps, more tasks |
| Make | 1,000 credits a month, 2 active scenarios | Core | More credits and scenarios |
| n8n | Community Edition is free to self-host | Cloud Starter, about €20 to €24 a month | Managed hosting; you pay the server if self-hosting |
| WhatsApp Business Platform | No free tier for templates | Per delivered template message | Marketing, utility, authentication messages; check Meta's rate card |
| Voice agents (Vapi, Retell) | Trial credits | Usage, roughly $0.07 to $0.30 a minute all-in | Speech-to-text, AI model, voice, phone line |
13.5 What you need before you start
Hardware and software: any laptop from the last five years works. For Claude Code you also need a terminal (built into macOS and Windows) and, for some installs, Node.js from nodejs.org. Nothing in this lab needs a powerful computer, because the AI runs in the cloud.
13.6 Accounts, API keys and safety
Many labs connect one app to another. There are three ways apps connect, and you will meet all three:
Five safety rules for every lab
- Never paste an API key into a chat, a document, a prompt, or frontend code. Keep keys in the tool's secret store (n8n credentials, Lovable or Replit secrets, Netlify environment variables).
- Use test data. Do not upload customer data, salary data or anything under NDA to a free AI plan. Check your company's AI policy first.
- Approve actions one by one until a workflow has run correctly at least ten times.
- Turn off what you no longer use. Disconnect connectors and delete unused API keys.
- Set spending limits on every paid API (Anthropic Console, OpenAI, Twilio, Vapi) before your first test.
13.7 A starter stack for three budgets
₹0 a month (learning)
- Claude Free
- Lovable Free
- Figma Starter
- Notion Free
- Zapier Free or n8n on your laptop
- Telegram bot for chat practice
- Voice agent trial credits
About $45 a month (serious PM)
- Claude Pro (includes Claude Code)
- Lovable Pro
- Figma and Notion free tiers
- n8n self-hosted on a small server
- Pay-as-you-go API with a $10 limit
Team or client work
- Claude Team or Max
- Lovable or Replit paid
- Figma Professional
- Notion Business
- n8n Cloud or Make Core
- WhatsApp BSP and a voice agent platform on usage billing
13.8 How the classes connect: one example project
The capstone that ties the lab together is a small business assistant for a hypothetical clinic. Each class adds one piece:
13.9 Exercises with model answers
Exercise 1 · Pick the tools
A D2C brand wants: a landing page with a waitlist form, an email to each signup, and a weekly Slack summary of signups. Which tools would you use, free first?
Model answer
Landing page and form: Lovable (free tier) or Claude artifacts for a quick static page. Email and Slack: Zapier (free covers a two-step "new form row → send email" Zap) plus a second Zap for the weekly summary, or one n8n workflow that does both. Summary text: an AI step (Claude via n8n, or Zapier's AI step). Paid only when signups exceed the free task limits.
Exercise 2 · Spot the risk
A teammate pasted the company's OpenAI API key into a Lovable prompt so the app "can talk to AI". What is wrong, and what should they do?
Model answer
The key is now in the chat history and may end up in frontend code where any visitor can read it. Revoke the key immediately, create a new one, and store it as a secret (Lovable Cloud or Supabase secrets) used only by a server-side function. Set a spending limit on the new key.
13.10 Key takeaways
- Choose the tool group by the job, and the tool by cost, integrations and team fit.
- Every Level 1 lab is free. Messages, calls and production hosting are what cost money.
- Keep secrets out of prompts and frontend code, use test data, and set spending limits.
- Connectors, API keys and MCP are the three ways tools talk to each other.
13.11 Check your understanding
Check your understanding
Which group would you use to change code inside an existing GitHub repository?
- AI assistants such as the Claude chat app
- Coding agents such as Claude Code
- Automation tools such as Zapier
- Voice agents
What usually costs money even when you are only testing a WhatsApp or phone bot with real people?
- Writing the prompt
- Designing the flow in Figma
- Delivered messages and call minutes
- Creating a Google account
Where should an API key for an automation live?
- In the prompt so the AI can see it
- In a shared Google Doc
- In the tool's credential or secret store
- In the website's JavaScript
13.12 Recommended resources
- Claude pricing · Lovable pricing · Replit pricing · Figma pricing
- Notion pricing · Zapier pricing · Make pricing · n8n pricing
- WhatsApp Business Platform pricing (Meta)
- Model Context Protocol — the standard behind Claude's connectors.