The Only AI Tools You Need: A Practical Guide to Smarter Workflows

In a world flooded with AI tools, the real productivity advantage doesn’t come from using more tools—it comes from using the right tool for the right job.

Jeff Su, a former Googler and well-known productivity creator, breaks this problem down elegantly: a small set of AI tools delivers 90% of real-world results if you understand their individual strengths. This guide distills that mental model and maps each tool to the workflows where it performs best.

This is Part 1, covering Everyday AI and Specialist AI. More advanced categories will follow in Part 2.

1. Everyday AI: General-Purpose Models with Distinct Superpowers

At first glance, general-purpose chatbots look interchangeable. In practice, each has a clear competitive moat that makes it superior for specific tasks.

🧠 ChatGPT — The Most Obedient Model

ChatGPT’s defining strength is instruction-following precision.

When given long, multi-step instructions with strict constraints, ChatGPT consistently executes every requirement without silently skipping steps. Other models may be equally intelligent, but they often “optimize” your instructions by deciding what matters and what doesn’t. ChatGPT doesn’t do that—it follows your rules exactly.

Why this matters

Complex prompts with dependencies
Hiring rubrics, evaluations, grading systems
SOP generation where missing one step breaks the workflow

Key Strength

Unwavering obedience and meticulous adherence to complex instructions.

Best Use Case

Any task with detailed rules, checklists, or interdependent constraints where correctness matters more than creativity.

🎥 Google Gemini — The Multimodality Master

Gemini dominates when your inputs are not just text.

It can natively process video, audio, images, PDFs, and slides—all in a single prompt. Combined with its massive context window, Gemini can ingest an entire meeting recording, a slide deck, and whiteboard photos simultaneously and synthesize them into structured outputs.

Real-world example Upload:

A 1-hour meeting video
A 20-slide presentation
A whiteboard snapshot

Ask:

“Summarize key decisions, action items, and draft a follow-up email.”

Only Gemini can do this in one shot.

Key Strength

Native multimodal understanding with an extremely large context window.

Best Use Case

Meeting summaries, SOP creation from walkthrough videos, training material synthesis, and mixed-media analysis.

✍️ Claude — The First-Draft Polisher

Claude excels at producing high-quality first drafts—both in code and writing.

1. Coding

Developers consistently report that Claude produces working code on the first attempt more reliably than other models. Whether it’s scripts, backend utilities, or data transformations, Claude minimizes back-and-forth debugging.

It also performs exceptionally well for:

Mermaid diagrams
Data visualizations
Turning static ideas into functional implementations

2. Writing & Style Matching

Claude shines when tone matters. Once you provide writing samples, it can replicate:

Corporate review language
Blog voice
YouTube or presentation scripts

The output usually requires far fewer revisions and sounds more “human.”

Key Strength

High-quality, near-final drafts for both code and written content.

Best Use Case

Writing production-ready code, polishing blog posts, refining scripts, or matching a specific writing voice.

🔁 How to Use Everyday AI Together

A highly effective workflow looks like this:

ChatGPT or Gemini → Ideation, research, outlines, raw structure
Claude → Final polishing, code generation, tone refinement

This “rough → refined” handoff alone can save hours per week.

2. Specialist AI: Precision Tools for Specific Jobs

These tools are not generalists. They are purpose-built to do one thing exceptionally well.

🔎 Perplexity — The Fast Information Fetcher

Perplexity is optimized for retrieval, not reasoning.

Think of it as a precision search engine rather than a chatbot. When you need:

A specific fact
An up-to-date answer
A sourced claim

Perplexity delivers faster and more reliably than general-purpose models.

Key Strength

Rapid, accurate, source-backed information retrieval.

Best Use Case

Fact-checking, quick research, confirmation queries, and targeted searches (e.g., site:reddit.com).

📚 NotebookLM — The Hallucination-Free Zone

NotebookLM only answers using your uploaded sources—nothing else.

This “walled garden” approach nearly eliminates hallucinations. If the information isn’t in your documents, NotebookLM simply won’t invent it.

Practical use cases

Fact-checking blog posts against research notes
Validating marketing claims before publishing
Ensuring video scripts align with source material

Important caveat

If your sources are wrong, NotebookLM will be confidently wrong.

Key Strength

Guaranteed source-grounded answers with zero external assumptions.

Best Use Case

High-stakes accuracy workflows where correctness is non-negotiable.

Final Thoughts: Choosing the Right AI Stack

You don’t need every AI tool—but you do need clarity.

ChatGPT → Precision and instruction-heavy tasks
Gemini → Multimodal and large-context workflows
Claude → High-quality first drafts and code
Perplexity → Fast, accurate information retrieval
NotebookLM → Source-locked verification and fact-checking

Jeff’s advice is practical:

Most people should master paid ChatGPT first.

If your workflow justifies it—and your budget allows—layering in these specialized tools can unlock a step-change in productivity.