How to: Become more productive (with AI tools)

Human intelligence remains one of the beautiful mysteries of the world: how our brains work, how they learn, how they innovate. As such, its no surprise that our ability to incorporate an entirely new form of intelligence into our workflows is a challenge. Using AI is a muscle that requires training and developing. It takes reps and upfront work before you can realize the full benefits.

When I first started using consumer AI tools like ChatGPT or Claude, I was both amazed at how well they worked and embarrassed and how inorganic it felt to incorporate them into my day-to-day work/life. I believed in their power and potential to disrupt, but I was struggling to “walk-the-walk” in actually adopting them.

Centaurs & Cyborgs

I have since learned that this is not an uncommon starting point. We are entering a new era of intelligence, one where work and thinking activities are divided between human and robot. To usher in this era, the robots need to get smarter, and are getting smarter every month. But the humans also need to adjust their how they work and think in order to accommodate.

I have been and continue to be inspired by Ethan Mollick's writing on how human work will be augmented and enhanced by AI in the future. He uses two terms to describe the ways humans and AI work together:

1. Cyborgs: Humans and AI work closely together, blending their efforts.

2. Centaurs: Humans and AI split tasks clearly. Each does what they're best at.

Within your day-to-day workflow, you need to find opportunities to be a cyborg (e.g., collaborating with AI to co-write reports or generate ideas together) and times to be a centaur (e.g., delegating routine data analysis to AI while you focus on the strategic takeaways and actions from the data). And I want to help you.

Step 1: Pick your tools

Today there isn’t a single tool for everything, you need to understand when to use each tool, their relative strengths and weaknesses

At risk of being overly bullish, I will say that its not hard to get $20 worth of value out of any of these tools once you commit to them. Starting out, almost everyone should pay for ChatGPT and the other ones can come later. Other tools I like: Playground for logo/content creation, Granola AI for note taking (both of which I use on the free tier).

Step 2: Adapt your workflows

Now that you’ve assembled the right models and apps to address problems in your day-to-day workflow, its time to adapt what you do to make AI a part of your processes.

If you haven’t already, you need to try voice as a modality for interacting with AI. And really invest some time in it. At first you will think you are talking to Siri, Alexa, Google - because thats how your brain is used to talking to machines. But if you invest, you will see how much thinking the LLM can do for you. The following video was the moment I fell in love with ChatGPT voice. Notice how incoherent her question is and how good the answer is (and this was in 2023):

(as mentioned in my last post, iPhone users should assign the action button on your phone to invoke ChatGPT voice).

I would also recommend finding ways to reduce the number of clicks/tapes/swipes required to use AI. Google search has dominated by minimizing the effort to execute a search (through chrome navigation bar, iPhone default search provider etc). AI is not as easy to ingrain yet, so you need to do it. I would recommend assigning hotkeys for invoking ChatGPT or Perplexity desktop apps on your computer. For example, if you use ChatGPT app for mac, by default Option + Space will invoke the chat window:

I’m a lunatic, so I bought this thing but I think that’s total overkill:

Step 3: Create Side Quests (until they become main quests)

This wave of technology innovation is by far the most user friendly one in history. All these models are coming to market with consumer apps built on top of them. The only prerequisite skill is curiosity. So I would recommend CREATING opportunities to experiment. These side quests are how you learn and can be a lot of fun at the same time. Some examples in my life:

My first blog post

The analysis of my first 600 chatGPT conversations was a total side project. And I learned a ton about how the data is structured within chatGPT, data volume limits, and its strengths and weaknesses in unstructured data classification

‘Favorite things’ book

I used ChatGPT to take a list of family’s favorite things and convert it into a rhyming song a la Sound of Music, then used Midjourney and ChatGPT DALL-E to illustrate it and playground.com to create the logo.

Fully functioning quiz app

Claude converted a long PDF my wife was trying to memorize into a react-based quiz app that kept score of her progress.

Podcast about my high school

Perplexity conducted in-depth research on the history of my high school, GPT o1 organized the information and NotebookLM converted that into a podcast.

If you are struggling to incorporate AI into your workflows, you are not dumb, you are not lazy, AI is not dumb, AI is generally not lazy. Keep at it.

AI Dad joke of the day:

Q: What happened when I asked AI to help with my kids' homework? // A: It aced the test but forgot to teach the kids!