Join the AI class
The workforce is dividing. Make sure you’re on the right side.
👋Hi, it’s Greg and Taylor. Welcome to our newsletter on how to make high-stakes professional and personal decisions in your 30s.
Read time: 10 minutes
I got lucky. I graduated from business school the year the internet was born. Down to my last $25K, I landed a consulting gig at Apple to work in their “new media” division. Apple was lost, but I was at least positioned in the biggest tech wave ever – the internet.
Sure, I took a risk, worked hard, blah blah blah. But I was also in the right place at the right time. I was lucky that a massive wave of opportunity and wealth creation was gathering steam. I positioned myself in that wave and rode it for the next few decades of my career.
Another career-defining wave is here: generative AI. AI is what the internet was 30 years ago. And just like with the internet, people who join “the AI class” now (early) will get higher compensation, faster promotions, less stress, and, maybe most importantly, time back.
If you’re ambitious, your biggest priority right now should be getting into the AI class.
Here’s our 10-minute pitch on why. I hope after reading it, you’ll (at least) be convinced to pay $20/month for GPT-4. If not, tell me why.
– Greg
Difficult to master + high impact = career-defining skill
Career-defining skills have two components:
They’re high impact. Workers with these skills are 10x more valuable because they accomplish better work faster than their peers. Therefore, these workers get better projects at work and have more impact, which usually turns into faster promotions and higher compensation.
They’re difficult to master. The harder a skill is to master, the fewer people will be willing to do it. Think coding in the late 90s or data science in the 2000s. High impact skills that are difficult to master create a talent deficit, which creates better opportunities for people with these skills.
AI will be a “career-defining” skill
Individuals using AI are more productive and produce better work
13 months ago, OpenAI released the first public version of ChatGPT. Since then, countless studies have emerged demonstrating AI’s impact on productivity and quality of work.
In July, MIT researchers recruited 450 knowledge workers and gave half access to ChatGPT. Users with access to AI completed writing tasks 40% faster than those not using AI, and the quality of their work increased 18%.
Also in July, Wharton and Cornell Tech researchers measured the speed and quality of product ideas generated by AI vs. MBA students. Not only did ChatGPT generate product ideas much more quickly than students – ChatGPT’s ideas were of higher quality as measured in purchase-intent surveys.
In October, HBS and Wharton partnered with BCG to study the impacts of AI on nearly 800 BCG consultants. Again, half were given access to AI. And again, those with access to AI were faster and better – completing consulting tasks 25% faster and at 40% better quality than those without access.
These results represent opportunity and risk for high performers.
Opportunity, because getting fluent in AI gives you another professional superpower.
Risk, because AI will reorder the workforce, and high performers who reject AI out of fear or skepticism could fall behind.
AI is difficult to master right now
There are three key reasons why learning and using AI isn’t easy right now.
1. The technology has flaws (e.g., hallucinations)
AI is a beta technology that’s been released as a general purpose technology. The UI of platforms like ChatGPT, Claude, or Perplexity isn’t always intuitive. These LLMs time out, have unexplained bugs, and work great some days and not others.
Today’s LLMs also occasionally “hallucinate” or make up answers. For most LLMs, 3-10% of responses are made up by the AI.
In 12-18 months, we’d expect to see the rate of hallucinations diminished or solved by LLMs. But in the meantime, AI users have to spot check their work, run queries through multiple LLMs, and still rely on humans to validate responses.
2. Its capabilities are changing fast
Every time you use AI, you’re using the worst AI you will ever use. Here’s one example from ChatGPT. In 13 months, ChatGPT has gone from simple text generation to real-time research, data analysis, code and image generation, and more.
Technologies that change fast are difficult and frustrating to master. In July, Taylor tried to use AI to write the first draft of her team’s performance reviews. It was terrible. In November, ChatGPT’s output was okay and Claude’s was great. But most people don’t have the patience or time to iterate to better results.
3. People are telling you it’s a waste of time
It’s easier to be a skeptic right now. Criticizing AI from a distance makes you feel smart, without the discomfort of learning something new or taking a risk. You’re likely hearing from your friends, “AI is a waste of time / going to take our jobs / going to ruin society” – or hearing from your company, “Don’t use AI on company time” or “They’re stealing our data.” It takes a lot of conviction to block out the naysayers and do it anyway.
What the AI class will look like
Because of this powerful combination – huge impact + hard to master – an AI class is emerging.
The AI class will be:
Role- and level-agnostic. Interns and CEOs will be in the AI class – in fact, more interns than CEOs will likely join the AI class since employees entering the workforce today will be “AI native.” 70% of Gen Zers already use AI.
Less anxious and more confident around AI. AI natives will see AI as a superpower that increases their value, rather than a threat that replaces them.
Less mired in busy work. The AI class will be more strategic because they’ll spend less time on drudge work – data analysis, document generation, email writing, and other time-consuming tasks.
Able to produce better ideas and strategies. The AI class will have a co-strategist and co-creative (AI) who can come up with endless ideas and gut-check their decisions.
5 steps to join the AI class
If you’re convinced – here’s how to get started.
1. Choose your go-to LLM
First, you need to choose which LLM you’ll use. We’d start by paying $20/month for access to ChatGPT-4 and using the free version of Claude by Anthropic. If you’re lucky, your company is rolling out Microsoft Copilot, giving you free access to GPT-4.
GPT-4 is the most advanced LLM – if you don’t pay for it, you won’t understand AI’s full capabilities. Claude is the best (in our experience) for creative, thought partner work. If you don’t use it, you won’t see the full potential of AI as a strategic copilot.
2. Set up your custom instructions
Next, set up your custom instructions (only available with ChatGPT right now). This feature allows you to give GPT context that it remembers across every interaction.
Think of this as onboarding a new teammate. Explain what GPT should know about you – your role, industry you work in, challenges you’re facing. Then explain how you want GPT to behave – its level of risk-aversion, how it should talk to you, and what it should say if it doesn’t know the answer.
3. Do five drills to get comfortable
Complete these five AI drills to get used to interacting with an LLM and understand its range of capabilities.
Upload a document. Attach a file and ask ChatGPT or Claude to “create an executive summary of this report.”
Generate an image. Ask ChatGPT to “create an image of a purple giraffe in the style of an impressionist painting.”
Analyze an image. Upload a picture of a houseplant from your house and ask ChatGPT, “What type of plant is this?”
Query the web. Ask ChatGPT, “How many users does ChatGPT have as of December 2023?”
Generate an idea. Ask Claude or ChatGPT, “Give me 5 ideas for high-protein, gluten-free dinner recipes.”
4. Find your use cases
AI can augment two types of tasks or workflows – high manpower and high brain power tasks.
High manpower tasks are repetitive – you perform them over and over again. Think generating templatized weekly emails, reviewing legal contracts, customer call prep, and weekly or monthly status updates.
High brain power tasks require high quality thinking, and can usually benefit from a second opinion. Think proposing a new initiative, deciding which product features to build next, or launching a new vertical.
To identify ways you could use AI, think of high manpower and high brain power tasks in your day-to-day life. Here’s an example for Taylor’s tasks as COO.
5. Invest 10 hours…then 15 minutes per day.
Spend 10 hours next week playing with Chat GPT-4 and Claude. Use the desktop version and mobile app. After 10 hours, evaluate what you learned. Look for:
No brainer use cases
Use cases that have promise, but aren’t working quite yet
Things you should keep doing yourself (for now)
Once you’ve identified 1-2 use cases that work, use AI every day for 15 minutes. Find the time to get confident with and fluent with AI. It’s the only way you’ll propel yourself into the AI class.
Our advice
Right now, 14,000 people in the US have “prompt writing” or “prompt engineering” listed as a skill on their LinkedIn profile. That’s 0.3% of the workforce. Next year, that percentage will be 10x. This will be the AI Class.
Here’s our experience after six months of using AI every day:
Every day, we outsource some research, quick analysis, or a data summary to ChatGPT or Claude – AI as an intern.
Once a week, we get valuable help from AI on a tough assignment or decision – AI as a thought partner.
Every few weeks, we convert a repetitive workflow to an AI automation using more advanced prompting or a custom GPT bot – AI as copilot.
Get your 10 hours under your belt, then invest another 15 minutes per day for 90 days (aka, join the AI class). That’s less than 2% of your working hours in 2024. If you experience even 10% productivity gains, that’s venture-like returns (3x). This is a career-defining wave. We want you on it.
Not convinced? Tell us what it would take.
To the next 10 years,
Greg & Taylor
P.S. If you’re ready to join the AI class, we’re launching a new 10-day AI Crash Course on Feb. 12 at Section. Use discount code PERSONALMATH for 20% off.
the real
d b