We do not need one AI to rule them all. We need the right AI for the right moment. After watching Jeff Su break down how he works with multiple models, I’m convinced the smartest move isn’t chasing the “best” tool. It’s designing a playbook. My take: use AI like a team, not a solo act, and you’ll get faster, cleaner results.
What Jeff Su Gets Right
Jeff Su, a product marketer known for practical workflows, argues that general chatbots aren’t interchangeable. He shows where each shines. His stance is refreshingly simple: match strengths to tasks, and stop expecting one model to do everything well.
“ChatGPT is the most obedient model… it follows each instruction to the letter.”
He proves it with complex prompts like hiring rubrics. ChatGPT checks every requirement. Others sometimes skip steps. That reliability matters when mistakes are costly.
“Gemini is able to process a massive amount of mixed media… video, audio, images, and text natively.”
Gemini isn’t always the top reasoner, but its giant context window and native handling of video and audio make it a workhorse for messy, real-world inputs.
“Claude’s first attempt is usually closer to done.”
For code and polished prose, Claude delivers strong first drafts. Jeff even got a working Go script on the first try. That saves time, and time wins.
“Perplexity is built for fetching.”
Perplexity isn’t for brainstorming. It’s for fast, accurate facts with sources. Think “grab-and-go” answers, not essays.
“NotebookLM… only answers from the sources you give it.”
When accuracy matters, this “walled garden” prevents guesswork. If your sources are solid, your outputs stay tight.
My Take As A Customer Loyalty Nerd
I build systems that create superfans. The throughline in Jeff’s workflow is trust. Trust comes from consistent, repeatable results. That’s what customers want, and it’s what leaders need from AI.
Obedience isn’t glamorous, but it’s gold. ChatGPT’s strength with multi-step instructions mirrors a great customer journey: no dropped balls. Gemini’s media intake mirrors how teams actually work: recordings, slides, screenshots, and notes. Claude’s style matching and working code speed up the last mile, where quality makes or breaks brand experience.
I agree with Jeff’s warning on Grok. If you don’t need real-time social data, skip it. Don’t add tools that don’t solve a clear problem. Tool bloat kills momentum.
Counterpoint And Balance
Some will say, “I don’t want three subscriptions.” Fair. Jeff even says most people can do great work with a single paid plan. I agree for many roles. But if you handle complex content and ship often, mixing tools pays off fast. The hidden cost isn’t the fee. It’s rework, missed steps, and bland outputs.
How To Apply This Now
Here’s a simple playbook I use after testing Jeff’s approach in my work with brands and teams.
- Complex, multi-step tasks with high stakes: start with ChatGPT.
- Anything with video, audio, or huge files: feed it to Gemini.
- First-draft copy or working code: finish with Claude.
- Fast facts with citations: check Perplexity.
- Source-locked accuracy checks: run through NotebookLM.
You can chain them, too. Draft ideas in ChatGPT or Gemini, sharpen with Claude, verify claims in Perplexity, and audit with NotebookLM. That sequence keeps speed without losing trust.
Standout Lessons From Jeff Su
His examples are sticky because they map to everyday work:
- Hiring rubric with a dozen rules: ChatGPT caught every requirement.
- SOP from a messy screen recording: Gemini turned video into a clean doc.
- Code on first try: Claude produced a working Go script for a bulk export.
- Trip planning vs. quick facts: use chatbots for plans, Perplexity for specifics.
- Pre-publish fact checks: NotebookLM flagged claims not backed by sources.
Final Thought
Tool loyalty is overrated; outcome loyalty is everything. Pick models for what they do best, build a repeatable flow, and ship with confidence. That’s how teams earn fans—inside the company and out.
Start this week. Pick one high-stakes task. Break it into stages. Assign each stage to the model that fits. Measure time saved and errors avoided. Then standardize it for your team. Your future self—and your customers—will thank you.
