03Choosing & Using AI Assistants
Claude, ChatGPT, Gemini, Copilot, open-source apps — how to pick the right assistant, use its power features, and set up a daily AI workflow that actually saves time.
What you'll learn
- Compare the major AI assistants and when to use each
- Use files, projects, web search and custom instructions effectively
- Understand free vs. paid vs. business plans and what they mean for your data
- Build a daily AI routine for freelance or small-business work
The assistant landscape
An AI assistant is a chat app wrapped around one or more LLMs, plus features like file upload, web search, memory, and connections to other apps. They change fast, so rather than chase every release, learn the categories and the features that matter.
| Assistant | Made by | Known for | Free option |
|---|---|---|---|
| Claude | Anthropic | Strong writing and reasoning, long documents, coding, projects, artifacts | Yes |
| ChatGPT | OpenAI | Broad feature set, image generation, voice, custom GPTs | Yes |
| Gemini | Google Workspace integration, long context, multimodal | Yes | |
| Microsoft Copilot | Microsoft | Integration with Windows and Microsoft 365 | Yes |
| NotebookLM | Research over your own sources, audio overviews | Yes | |
| Perplexity | Perplexity | Search-first answers with citations | Yes |
| Open WebUI / LM Studio | Open source | Run private assistants on your own hardware (Track 4) | Free |
Power features most people never use
File & image upload
Drop in PDFs, spreadsheets, screenshots, contracts. Ask questions about them instead of re-typing content.
Projects / workspaces
Save reusable instructions and reference files per client or topic, so every new chat starts with context.
Custom instructions & memory
Tell the assistant once who you are, your business, tone and preferences.
Web search
For anything current — prices, laws, news, competitors. Always check the cited sources.
Code execution / analysis
Upload a CSV and ask for totals, charts or cleanup. The model writes and runs code for exact results.
Connectors
Link Gmail, Calendar, Drive, Notion, Slack and more so the assistant can read and act on your real data.
Free vs. paid vs. business: what happens to your data
This is the part most people skip, and it matters when you handle client information.
| Plan type | Typical data handling | Use for |
|---|---|---|
| Free / personal | Conversations may be used to improve models unless you opt out (check each provider's settings). | Learning, personal tasks, non-sensitive drafts |
| Paid personal | More usage and features; training settings vary — review them. | Daily freelance work without confidential data |
| Team / Business / Enterprise | Usually excluded from training by default, with admin controls and data agreements. | Client data, company documents, teams |
| API | Generally not used for training by default at the major providers; retention policies apply. | Automations and apps you build |
| Local models | Data never leaves your machine. | The most sensitive work (Track 4) |
Set up your assistant properly (15 minutes, once)
Write your “about me” instructions
Most assistants have a settings area for custom instructions or a profile. Paste something like:
custom-instructions.txtprompt I'm a freelance web designer based in Iloilo, Philippines. My clients are small businesses (cafés, clinics, resorts). I bill in PHP and USD. When you help me: - Be concise; use bullets for lists and tables for comparisons. - Write client-facing text in warm, professional English (Taglish only if I ask). - If a fact is uncertain or time-sensitive (prices, laws), say so. - Ask me a clarifying question when a request is ambiguous.Create a project per client or recurring job
Add the client's brand guide, service list, past proposals and FAQs as project files. Add project instructions like “You are the account assistant for Kapé Iloilo…”.
Save your best prompts
Keep a prompt library (Notes, Notion, or the Toolkit's prompt library on this site). Good prompts are reusable assets.
Turn on the connectors you'll really use
Calendar and email are the most valuable for day-to-day admin. Start with read access and add write actions once you trust the setup.
A daily AI routine for freelancers
| When | Task | Example prompt |
|---|---|---|
| Morning | Inbox triage | “Here are my unread emails. Group them into urgent / reply today / FYI, and draft replies for the urgent ones.” |
| Before calls | Prep | “Summarize everything in this project about the client's goals and open questions. Suggest 5 questions to ask.” |
| After calls | Notes → actions | “From this transcript, list decisions, action items with owners, and a follow-up email.” |
| Deep work | First drafts | “Draft a proposal using my template and this discovery-call summary.” |
| End of day | Admin | “Draft polite payment reminders for these overdue invoices.” |
Key takeaways
- Pick one main assistant and learn it deeply; skills transfer across tools.
- Files, projects, custom instructions, search and connectors are where the real productivity is.
- Know which plan you're on and how it treats data before using client information.
- Save reusable instructions and prompts — they compound over time.
Knowledge check
0 / 3Q1Where should confidential client documents go first?
Business plans and local models offer stronger data protections; always check contracts and the law.
Q2What's the main benefit of a project/workspace feature?
Projects keep persistent context so each new chat starts informed.
Q3You need today's exchange rate. Which feature should you use?
Current information requires search or tools; the model's training data is dated.