Skip to content

Example: maximize my AI subscription

Last verified: 2026-05-06 · Time: 30–45 minutes · No API key required

You already pay for an AI. This example finds the highest-leverage features you haven't tried yet — and gets you to use one of them this week.

It's also the example that reveals when to switch tools.

Step 0 — Pick your product

Whichever AI you spend the most time in. If you have multiple paid products, pick the one you opened today.

Step 1 — Build your Portable AI Profile

If you don't have one, do this first. It pays back in every other workflow:

Now every chat in that product knows your tone, role, and constraints.

Step 2 — Run the audit

Paste this in your primary AI:

Audit how I'm using you. Based on this conversation and what you remember about me,
list 5 features in <ChatGPT/Claude/Gemini/Grok/Perplexity/Copilot — pick yours>
that I have probably not used yet, ranked by leverage for my work.

For each feature:
- Name it as it appears in the UI.
- One sentence on what it does.
- One sentence on why it would help me specifically.
- One sentence on the easiest first use.
- Plan availability (Free / Paid / Team / Developer).

If a feature is gated by a plan I don't have, say so and suggest the closest free fallback.
Do not invent features. If unsure, say so.

Read the output. Cross-check against the Capability map and your AI's Mastery page.

Step 3 — Pick one. Use it this week.

The point is using one new feature, not building a perfect plan.

Common wins by product (drift risk: medium):

ChatGPT

  • Custom Instructions (Free) → set tone + role + "avoid" list.
  • Memory (Sub) → tell it your role and a few constraints.
  • Projects (Sub) → group all chats about one topic; instructions follow you.
  • Custom GPTs (Sub) → publish your most-repeated workflow.
  • Tasks (Sub, where rolled out) → schedule a daily/weekly prompt. Manual-test first.
  • Canvas → for any document longer than half a page.

Claude

  • Profile preferences (Sub).
  • Projects + Project Knowledge (Sub).
  • Artifacts (Free / Sub) → for code, docs, web mockups.
  • Skills (Sub, growing) → package a process as a SKILL.md.
  • Cowork (Sub, where rolled out) → schedule a recurring task.

Gemini

  • Saved info (Sub).
  • Gems (Sub) → reusable assistants.
  • Workspace integration (Sub) → Drive / Calendar / Mail context.
  • Deep Research (Sub) → use it for one question a week.
  • Scheduled actions (Adv, where rolled out).

Grok

  • Personalization (Sub, where rolled out).
  • Model picker → run the same question through two variants; learn which you'll trust.
  • Grok on X → quick second opinions on threads you're already reading.

Perplexity

  • Spaces (Sub) → one Space per topic, with sources.
  • Profile + AI tone (Sub).
  • Pages (Sub) → convert research into shareable articles.
  • Deep Research → for citation-heavy questions.
  • Comet (Sub) → AI-native browser, where rolled out.

GitHub Copilot

  • .github/copilot-instructions.md (Free) → repo-aware prompt.
  • Selection-based chat (Free) → highlight code, refactor selection only.
  • Slash-prompts in .github/prompts/ (Free) → reusable patterns per repo.
  • Coding agent (Paid) → assign an issue. Review the PR like a junior dev's.
  • Copilot Extensions (Team / Ent).

Step 4 — Decide whether to switch tools

Sometimes the highest-leverage move is a different AI. Use this rule of thumb:

If your top need is And your current AI is weak there Consider
Citations ChatGPT / Claude / Gemini Perplexity
Long documents ChatGPT / Gemini / Grok Claude
Sharing assistants with a team Claude / Grok ChatGPT / Gemini
Email & calendar Grok / Perplexity Gemini (Workspace) / ChatGPT
Code in a repo Anything except Copilot GitHub Copilot

See the Capability map for the full version.

Step 5 — Lock in the win

End-of-week:

  1. Did you use the new feature at least three times?
  2. Did you save the prompt or assistant somewhere you'll find it?
  3. Add a one-line note to your portable AI profile under "Recent changes."

If yes to all three, you've raised your floor.

Layered version

This example is itself a workflow. You can package the audit prompt as a Project / Gem / Custom GPT and re-run it quarterly.

What good output looks like

  • 5 features that are real and named correctly.
  • Each one is plausibly the next thing for you.
  • At least one is in your current plan; at least one is in a higher plan with a fallback.

If your plan doesn't have a feature the audit suggests

That's the fallback question. Read the Capability map row for that capability. Use the manual workflow until the upgrade is worth it.

See also