The Working Class in the Early AI Era (Part 2)

Before the Tools, Build the Foundations

In Part 1, I asked how we could bring AI into our working lives without giving up the stability we already have.

Before we get into specific tasks, I want to take a step back. Learning to work with AI starts with more than choosing a tool. It starts with how we approach something unfamiliar.

Starting something new is difficult. I’m writing this at 32, and I sometimes wonder whether starting again in ten years will feel even harder. I don’t know the answer. But it reminds me to be kinder to myself while I’m learning.

You won’t become an expert tomorrow. You might struggle with something that looks easy to someone else. You might even discover a different interest along the way. That doesn’t mean the effort was wasted.

AI adds another kind of friction: the tools keep changing.

You can spend weeks learning a workflow, only to see a new release make it much simpler. In the AI communities I follow, people move to a new provider, then move back when another update arrives. It can feel like an endless race to keep up.

But a better model doesn’t erase what you’ve learned. Some tool-specific tricks may become less useful. Your ability to define a problem, provide context, check an answer, and decide what to do next still matters.

For me, three foundations make this learning process more sustainable.

1. Stay curious without chasing every update

Start by understanding the tool you already have. Where does it help? Where does it struggle? What kind of work still needs your judgment?

Stay open to alternatives, but give yourself enough time to learn through actual use. You don’t need to switch tools every time someone announces a new benchmark winner.

A useful question is: “Does this help me do my work better?”

2. Manage the work, not just the conversation

A little technical understanding helps us set realistic expectations. Generative AI can produce text, images, audio, and other content. Large language models, or LLMs, are one important part of that landscape. They generate language by predicting tokens, which can be words or parts of words, based on context.

Fluent output, however, is not proof of accuracy.

That is why I treat an AI-assisted task like a small project. What are we trying to achieve? What information do we have? What are the constraints? How will we know whether the result is good enough?

I find the SMART framework useful here: make the objective specific, measurable, achievable, relevant, and time-bound.

Instead of asking, “Help me improve this report,” I could ask for a one-page management summary that highlights three issues, uses only the supplied data, and flags missing information rather than guessing.

Write those expectations down. A simple document is enough. You may see people use Markdown files ending in .md to organize project instructions. The value comes from clear, reusable context, not the file extension itself. The AI tool also needs access to those instructions when doing the work.

For workplace tasks, use approved tools and follow your company’s rules on confidential information.

3. Learn to give direction and check the result

Talking naturally to AI is a good starting point. You don’t need to sound like a programmer. But a useful conversation still needs a clear purpose.

The learning resource linked below introduces four approaches:

• PTFC: Persona, Task, Format, Context. Clarify the perspective, the task, the output you need, and the background that matters.

• 3C: Concise, Clear, Consistent. Keep instructions understandable and avoid conflicting requirements.

• Supercharging: Add useful directions, such as “flag your assumptions,” “compare alternatives,” or “check this against the criteria.” Treat these as practical instructions, not magic phrases.

• Prompt chaining: Break a larger task into connected steps. For example, identify the issues, evaluate possible actions, then draft the summary. Review each step before moving on.

Good prompting is only part of the process. Check important facts, inspect calculations, and compare the result with your original objective. Ask for sources where appropriate, and verify that they support the answer.

The responsibility for what we use remains with us.

Keep the foundations, even when the tools change

My priorities are simple: build a mindset that lets you keep learning, give the work a clear structure, and communicate with enough clarity to evaluate what comes back.

You don’t need to master every tool. You need a way of learning and working that you can keep improving.

In Part 3, we’ll break down routine tasks and explore where generative AI can create practical value in our working day.

Explore further

For a deeper introduction, explore the accompanying NotebookLM resource on AI foundations, prompting frameworks, practical examples, and responsible use. It is an AI-generated learning resource, so follow its source references and check important claims as you read.

https://notebook.google.com/notebook/a45f88d8-6f02-44f8-96c3-cb00dde1701f/artifact/6d2d8e55-4050-4302-83c0-9bd041579737