174: What If?
A 2025 guide to understanding and gaining advantage with AI
First. Here’s a thing you could do. Let’s assume the world of finance is a blurry concept, at best. Then, you stumble upon the writer Matt Levine and his Bloomberg newsletter, Money Stuff. In brief order, arcane concepts become manageable. Derivatives become not just clear, but oddly fascinating. And you find yourself nodding along to the idea that, yes, everything is securities fraud. 🎂 Money Stuff celebrates its tenth anniversary this week. If you haven’t, you should subscribe.
In the past few weeks I’ve had the good fortune to talk about AI with diverse groups of people—in corporate gigs, academia, and inside agencies. I think Ben Evan’s chart above continues to ring true. Here at the onset of 2025, more people are still waiting on the sidelines than actively (daily) in the AI game.
What might it take to shift some of you from hesitant to enthusiast?
If you’re deeply immersed, today’s post might seem a bit remedial. Feel free to skip to the Updates below.
To begin, I want to recast a philosophical concept. Then I’ll propose a handful of ways anyone can easily incorporate AI into daily habits, gain confidence, and gain advantage.
You’re in the What If? business now
There are at least two truths about using LLMs—1) they are tools which can teach you how to use them, and 2) they function unlike previous software—which suggest a grounding philosophy. Each of us is now in the “What If?” business more than ever before. Every instance of uncertainty, curiosity, or vacillation has a dance partner, eager to move you forward.
And you don’t need permission.
Yes, the chatbot interface is maddening. What do I type? You type anything, even, “I’m not sure where to begin.” That’s enough to get started.
The biggest hurdle to get over when it comes to leveraging AI is your mental model of how computers function, and how an AI might help you. That’s why a What If? philosophy is critical. The point isn’t to know, beforehand, how it will turn out. That’s the deterministic, old computing stereotype hard at work. The point is to not know, to allow a probabilistic methodology to translate, contextualize, re-imagine, summarize, interrogate, and otherwise converse with you along a journey. Even if the task is simple.
Start with What If?
You won’t get into trouble. I promise.
Before we dive in: Anthropic’s Claude, OpenAI’s ChatGPT, Google’s Gemini, and Meta’s AI all offer a free desktop and mobile experience. You should sign up for at least two, and install their apps on your smartphone.
Then keep at least one model open, at the ready, all the time. And use it.
Fluency is habitual. (Duolingo is laughing at me right now.) Here are a handful of experiments to try on either desktop of mobile, in any of the major frontier models I listed above. Try each exercise multiple times, but change the topic or context. The point is to gain familiarity with each model’s user experience, and continued exposure to a probabilistic (i.e. uncertain) versus deterministic (i.e. certain) approach to engaging with AI software.
💬 If you have 10 minutes a day
Get used to using AI as a conversational thinking partner
1️⃣ Ask your AI a basic question (e.g., “How do I improve my marketing strategy?”). After it replies, ask it to suggest a better question to ask. After each response, keep asking for a better question to ask. See where the insights lead. This will help you reap the “thinking partner” benefit.
2️⃣ Take a strong opinion you have (a subject you know very well), and ask your AI to argue the opposite side convincingly. That’s the prompt: “I believe X about Y. Argue the opposite convincingly.” Try it with something cultural, or something from your industry. You’re learning how AI’s are conversational.
3️⃣ As your AI to devise a 10-minutes per day tutoring plan for a subject you don’t know but want to. Revisit the prompt every day and continue the curriculum. You’ll learn how prompts can become conversations over days and weeks.
🤹♂️ If you have 30 minutes a day
Get used to solving or re-imagining diverse, multi-modal tasks with AI
4️⃣ Upload an image of a PowerPoint slide, social media post, or webpage design (if desktop), or take a picture of a street sign, store display, or poster (if mobile) and ask, “How can I improve this for clarity and impact?” Don’t just accept its initial response either. Instead, ask “why did you recommend that?” Bonus: Upload two similar images (e.g., versions of a design, before-and-after shots, if desktop) or take two pictures (if mobile) then ask, “What changed between these?” Or, “Which of these is [attribute you prefer]?”
5️⃣ (Typically for desktop) Upload a dense report, legal document, or academic paper and prompt, “Summarize this in simple terms.” Then, ask it to rewrite the summary as if explaining to an 8-year-old, or board of trustees, or any persona. If you have a spreadsheet, P&L or finance doc, ask your AI to convert it into a dashboard. Then ask it to run scenarios based on what it can see in the data, i.e. “How can we reduce expenses?”
6️⃣ Upload or take a picture of a product, building, or machine and ask, “How does this work?” Tell your AI you’ve been asked to use the product, sell the building or operate the machine—ask it to write you a multi-part training plan. Then work through each step of the plan with your AI, asking it for feedback to your progress, potentially giving it more pictures of the product, building, or machine.
🕒 If you can spare an hour a day
Learn to build a productive relationship with AI for complex, “down the rabbit hole” assignments
7️⃣ (This is #3, but evolved) Pick a challenging topic you’ve wanted to understand deeply. Ask your AI for a step-by-step learning path with beginner, intermediate, and expert-level content. Insist on links to videos, articles and blog posts as part of each step. Then begin Step 1. After you review the content your AI recommends, as it to devise a quiz to test your knowledge. Ask it to grade your responses, and coach improvement. Continue through the remaining Steps.
8️⃣ Upload or paste your resume, portfolio, or LinkedIn profile URL. Ask your AI to, “Analyze my experience and suggest three future career paths I should explore. Make sure one is surprising yet relevant.” For any path you want to pursue, ask your AI to suggest a 30-day plan for improving your experience to make your resume/portfolio more relevant and appealing. With each recommendation, ask it what success looks like. If you pursue an approach, make updates to your resume/portfolio and upload it to the original prompt and ask if the new version is better aligned with the goal.
9️⃣ Tell your mobile AI you’re going on a walk together, and you’re going to take ten pictures along the way and add them to the same prompt thread. And the AI is going to write a novel based on the pictures you provide. Give it a word count per picture. You decide the genre, etc. Make the pictures random (architecture, street art, historical markers, specific vehicles, people in noticeable roles). Consider asking your AI to connect the images to history, pop culture, or another theme. See how your choice in images builds on and effects the narrative.
I hope you’ll try at least one of these, or the collection of examples helps get you free from hesitation and into trial and experimentation. Remember, an AI can teach you how to use it more effectively. If you get stuck, or the outcome isn’t actionable—just say so. “Hey, this isn’t working. How can I prompt you to get better results?” And your AI will generate something.
And please let me know what works, or how you’ve found motivation to leverage AI more consistently each day.
🙋🏼♂️ If you want more direction on where to start with AI, I suggest two recent posts.
Wharton Associate Professor
wrote his most recent “An Updated Opinionated Guide” to working with AIProduct VP
Jones recently published his “Complete Q1 2025 AI Productivity Stack” for working with AI
You could also read Mollick’s book, Co-Intelligence. Like his newsletter, it’s chock full of actionable wisdom. If you’re more interested in the bigger, philosophical picture, I recommend Mustafa Suleyman’s The Coming Wave (he’s a co-founder of Google’s Deepmind and now the CEO of Microsoft AI).
AI+Creativity Update
🏈 The Super Bowl arrives this weekend. I’ll be watching the ads, and Kendrick Lamar at halftime. YouTube’s aggregating all the commercials related to the big game.
📺 Curiosity+Courage podcast guest Ashley Rutstein has a keen take on the phenomenon of big game teaser ads.
🤖 OpenAI (finally) has a coherent, albeit expected, new brand identity. Wallpaper has the story. But I’m confused how this is a “first ever rebrand.”
🤔 Gordon Young, Editor-In-Chief of The Drum, weaves together three recent interviews with Sir Martin Sorrell about the state of agencies, advertising, and the impact of AI. The money quote:
“Meta’s Advantage+ is now running at a $20bn-a-year clip. That’s small and mid-sized businesses going direct to an automated system for content, media planning, and buying, all in one. No agencies, no planners, no strategists. Now, what happens when large clients start to do the same?”
🗝️ Another favorite writer of mine, Bob Lefsetz, is unafraid to illuminate the difficulty of being an artist. In The Magic, Bob makes it plain:
“Somewhere along the line it became conventional wisdom that everybody could be an artist, if they just tried hard enough.
This is patently untrue.
Artistry is a calling. A walk into the wilderness. And it’s the intangibles that put you over the top.”
As much as AI might unlock or empower or instigate creativity, being an artist is still hard work.