Search for...
Light Dark Auto
AI Posts

AI

First things first: If you're looking to stay ahead in the world of AI without a technical background, I recommend signing up for my free AI Toolkit (a 9-part email course covering the essential tools, techniques, and mental models) and watching my AI playlist on YouTube.

The AI world can feel like a wild jungle for the average working professional: full of hype, confusion, and constant change. It's easy to feel lost, and even easier to waste time chasing the wrong tools.

That being said, here are a few principles (and their corresponding implications) that I believe will stand the test of time:

Principle 1: Start with the Problem, Not the Tool

With so many new AI tools (and features) emerging daily, it's tempting to get caught up in the hype and start trying them all out. But this is a mistake.

  • Case in point: When OpenAI announced Custom GPTs, they convinced many developers to spend time creating GPTs by making a big deal about monetization via the GPT Store. And yet, very little progress has been made.

Here's what I recommend

The next time you come across a flashy tool or a new feature update, ask yourself these three questions:

  1. What problem does this tool solve?
  2. Who's currently using it? Are they in my peer group?
  3. Where would it fit within my current workflow?

If you find that the answers don't align with your core needs, it's likely a distraction rather than a value-add.

💡
Put simply: Once you've identified a problem that AI can help you solve, then you can start looking for the right tool for the job. In my experience, the AI tools from a handful of major companies are more than enough for the majority of our needs.

My AI toolkit

I use five tools on a daily basis, each with a distinct superpower:

  • Best for quality first drafts: Claude. My go-to for writing, analysis, and anything that requires nuance. I also use Claude Cowork to run my entire business operations (more on that below).
  • Best for following instructions: ChatGPT. When I need the AI to do exactly what I say without improvising, ChatGPT is the most obedient.
  • Best for multimodality: Google Gemini. Handles images, video, and Google Workspace integration better than anyone else. Want to test this yourself? Upload a YouTube video and ask Gemini to summarize it with timestamps.
  • Best for research: Perplexity. Fast, cited answers with real sources. My default when I need to look something up.
  • Best for grounded answers: NotebookLM. Upload your own documents and get answers that stay within what those documents actually say (no hallucinations, no made-up facts).

Edge use cases: Midjourney and Google's ImageFX for image generation, ElevenLabs for voice, Runway for video.

💡
Should you pay for these tools? I pay for all five daily drivers. The free tiers are fine for testing, but the paid versions are meaningfully better for real work. Unless you can't afford it (in which case, DeepSeek and Qwen are solid free alternatives).

Principle 2: Start with Augmentation, Graduate to Systems

If you spend any time on Twitter or LinkedIn, you've probably noticed the AI industry jumped from "chatbots" straight to "autonomous agents" and skipped the middle step where the value is: AI workflows. According to McKinsey, fewer than 10% of organizations have scaled true AI agents, because fully autonomous AI still faces massive hurdles like data security. We're looking at the decade of agents, not the year.

Start with augmentation. Find one repetitive task in your week and build a simple AI-assisted workflow around it. I used to spend 30+ minutes every week writing project recap summaries. I spent a one-time 30 minutes building a prompt that turns raw recap emails into concise, formatted summaries. The prompt does the heavy lifting, I do the quality check.

Then graduate to systems. Once you've built a few of these, they start connecting. Your meeting notes prompt feeds into your action items prompt, which feeds into your weekly summary prompt. Three separate prompts become a connected system that handles an entire workstream.

I've taken this to the extreme with Claude Cowork, which I use to run my entire business: email drafting, content production, project management, financial operations, and this website. Each of those started as a single prompt for a single task. Over time, they connected into a workspace that runs on accumulated context, not one-off conversations. If you're curious what that looks like, the Cowork Toolkit is a free, hands-on introduction.

💡
Pick one recurring deliverable you produce (a weekly report, a client update, a status email). Break it into steps. Let AI handle the predictable parts. Keep yourself in the loop for the judgment calls.

Principle 3: The Learning Loop

The AI landscape changes every week. It's tempting to try to keep up with everything: every new tool, every model update, every trending tweet. This is a recipe for burnout, not productivity. (Trust me, I've tried.)

The difference between "I know about AI" and "I'm good with AI" comes down to one thing: whether you actually do something with what you learn. Ask 10 colleagues what they've learned about AI this month. Most will mention something they read or heard about. Very few will tell you something they actually tried.

I use a simple system I call the Learning Loop. It takes about 30 minutes a week and has three steps:

Step 1: Consume (10 minutes a day)

You don't need to read everything. You need one reliable daily source that gives you the headlines in 5 minutes. I subscribe to one AI-focused newsletter (just one) and skim it over morning coffee. For a weekly deep-dive, Ethan Mollick's One Useful Thing is my go-to for understanding what new developments actually mean for how we work.

The win condition for consumption isn't "I read 25 articles today." It's "I found one thing worth trying."

Step 2: Take Action (30 minutes a week)

This is where most people fail. They consume plenty of AI content but never do anything with it. Block 30 minutes once a week (I do Saturday mornings) and pick one thing:

  • Read a headline about a new AI capability? Test the claim yourself. Open ChatGPT or Claude and see if it actually works as advertised.
  • Discover a new tool? Try it on a real task from your actual job, not a toy example. Give it 15 minutes. If it doesn't click, move on.
  • Learn a new prompting technique? Apply it to a prompt you've already saved.
💡
If you consume one piece of in-depth AI content a week and take action on one thing, you're ahead of 95% of your peers. That's not a motivational line. It's math: almost nobody does this.

Step 3: Share what you learned

This one is optional but powerful. When you explain something you learned to a colleague, a friend, or even in a quick note to yourself, you discover the gaps in your own understanding. You don't need to write a blog post. Just tell one person about one thing you tried this week: "Hey, I tested [tool] for [use case] and here's what happened." That's enough.

If you're looking for specific, actionable AI tutorials, these are my most-read posts:

Go deeper

If you want structured, step-by-step learning (beyond the free resources above), here's what I've built:

  • AI Toolkit (free): A 9-part email course covering the essential AI tools, techniques, and mental models for non-technical professionals.
  • Cowork Toolkit (free): A hands-on introduction to building your own AI-powered workspace with Claude Cowork.
  • AI Systems Academy (coming soon): A comprehensive course for non-technical professionals who want to build AI skills they'll actually use at work.