In 2018, I gave my two younger brothers some advice about their careers.
One was figuring out what to study. One was already in school but was thinking about changing direction. And I told both of them: go into software and go into accounting.
One became a software engineer. One became a CPA.
At the time, that was genuinely good advice. Software was hard to build — it required real skills, real training, and there was strong demand because most businesses could not build anything without hiring someone who actually knew how to code. Accounting had a massive moat: years of study, regulatory knowledge, and professional certifications. Both fields had high barriers to entry and strong job security.
Today, I would give them different advice.
Not because those fields are dying. They are not. But because the moat has shifted — and if you do not understand how, you are going to be caught off guard.
What Changed
The CEO of Anthropic recently said something that stopped a lot of people in their tracks: most of the software they build internally at Anthropic is now coded by AI, not by software engineers.
That is not a dig at engineers. Anthropic is still one of the most engineering-intensive companies on the planet. But it tells you something important about where the leverage has moved.
Software that used to cost a company $50,000 and three months of a developer time can now be built in days. The barrier to entry for building software — apps, tools, automations — is collapsing. Which means the moat that used to come from knowing how to code is narrowing.
Same thing is happening in accounting. AI tools are handling analysis, categorization, reconciliation, and reporting at a level that would have taken a trained accountant significant time. The technical skills that built the moat are getting easier to acquire — or skip entirely.
The Moat Has Moved, Not Disappeared
Here is what I want to be clear about: this is not a story about AI replacing people.
I work with a lot of professionals across different industries, and the ones I see thriving right now are not the ones who abandoned their domain expertise. They are the ones who took their domain expertise and layered AI fluency on top of it.
A finance professional who deeply understands financial statements and knows how to query AI tools to find patterns, flag anomalies, or model scenarios — that person is doing the work of three people.
A software developer who can architect systems and use AI to generate, test, and iterate on code faster — that person ships things faster than a whole team did five years ago.
The domain knowledge still matters. In some cases, it matters more than ever — because you need to know when the AI output is wrong. But the skill that sits on top of all of it is how well you can work with AI.
The Three Levels I Teach
When I work with people on this, I use a framework I call AI Fluency Levels.
Level one is AI Assisted — you are using AI tools like ChatGPT for one-off tasks. Prompting, generating, reviewing. This is where most people are.
Level two is AI Workflows — you have started connecting tools and building repeatable processes. Instead of copying and pasting between tools, you have systems that do it for you.
Level three is Building Agents — you are deploying AI agents that run autonomously and handle whole workflows without your involvement. This is where the biggest productivity multipliers live.
Most people stop at level one. The gap between level one and level two is where the moat starts to form.
What I Would Tell My Brothers Now
If I sat down with them today, I would say the same things I said in 2018 — learn the fundamentals deeply, get the credentials, build real expertise in your field.
But I would add one more thing.
Get good at working with AI. Not just using it — working with it. Building with it. Understanding what it is good at and where it needs you. Managing a system where you are directing AI to do the repetitive, automatable parts so you can focus on the judgment calls.
That is the skill I did not know to teach in 2018. It is the one that matters most right now.
The people I watch fall behind are not the ones who do not know their field. They are the ones who know their field but are doing everything manually while others stack AI on top of the same domain knowledge and produce at a completely different speed.
The moat has moved. The question is whether you move with it.
