Beat AI or Let AI Beat You

AI feels like an enemy. From a certain angle, it is. But mostly it’s just scary, the same way the internet was scary back in the day, and the same way personal computers were scary before that.

It helps to look at what already happened.

In the late 70s and 80s, when spreadsheet software showed up (e.g. Excel), an entire job category got hit. Bookkeepers and accounting clerks, the people who manually balanced ledgers and recalculated columns by hand, started disappearing. Their numbers in the US dropped from around 2 million in 1987 to about 1.5 million by 2000. The software did in seconds what used to take a day. 

But here’s the part most people skip: the total number of finance jobs didn’t shrink. Roles like financial analyst, management analyst, and auditor grew over the

same period, from about 600,000 to 1.5 million. Same humans, different work. The boring mechanical part went away, and the thinking part expanded. The people who got crushed were the ones who refused to learn the new tool. The ones who adapted ended up doing more interesting work for more money.

We’re in a similar moment again. Just bigger, faster, and louder.

The pace is the first problem. Models, frameworks, tools, agents, protocols. Something new ships every week and the old playbook breaks. Keeping up is mentally heavy.

The second problem is that the world is small now. Someone in a third-world country is scrolling reels about YC (Y Combinator) startups in San Francisco while drinking their morning coffee. Comparison used to be local, and now it’s global. That weight compounds.

The third problem is workplace pressure. Some employers are pushing AI usage hard. A few are tracking it. Some are even estimating token usage per employee. That is a real thing now. Whether or not you think it is a good idea, it tells you where the wind is blowing.

So is this the end of employment? Probably not. But it is the end of a certain kind of comfort.

The bookkeeper who learned Excel didn’t get replaced. The bookkeeper who didn’t learn Excel did.

The same logic holds now. AI by itself is not the threat. The threat is another person who has your domain knowledge and also knows how to wield AI. They will ship faster, write better, scale wider, and cost the same. Maybe less.

Domain knowledge is still where the value lives. AI doesn’t replace the engineer who knows why a Kerberos ticket gets refused, or the doctor who knows which question to ask the patient first. It amplifies them. AI is a multiplier on what you already know. A multiplier on zero is still zero.

A few things become real that weren’t real before. 

Boring, repetitive tasks can finally be automated by people who couldn’t write code last year. A marketer can build a small tool to clean their data. A coder can put together a landing page that actually converts. That crossover is real, and it is useful. But the real edge still comes from doubling down on your main thing. You don’t want to be mad at everything. You want to be excellent at one thing and assisted at the rest. 

Automation also gets smarter. Old automation was rule-based and brittle. New automation can reason, branch, and make informed decisions. The agentic era is built on this idea, AI that can take steps on its own to reach a goal. This is going to reshape how most workflows look in the next few years. 

Learning is faster too. AI can read papers, summarize them, build exercises, generate labs, and answer follow-up questions until something clicks. As someone who teaches, I can tell you the gap between a curious learner with AI and one without is bigger than it has ever been.

People treat AI like a magic black box. It is not. It is a stack of skills you can actually study. 

Prompting is a skill. Context engineering is a skill. Memory, tool use, agent design, evaluation, retrieval, fine-tuning. All skills. The people who treat AI as “the thing you type questions into” will plateau. The people who treat it like a discipline will keep growing. 

This is the new gap, and it is widening fast.

Some balance is needed. 

AI is great at many things. It is also wrong often, slow sometimes, and unnecessary in plenty of cases. A grep command beats a chatbot for finding a string in a log. A five-line bash script beats a prompt for renaming a thousand files. Manual intervention often wins on speed and certainty. Use the right tool, not the trendy one. 

Over-reliance is the bigger trap. We have already seen incidents where an AI agent with too many permissions wiped production data. Giving AI uncontrolled access to your infrastructure, your money, or your decisions is a real risk. It is a security problem too. 

There is a personal version of this risk. If you outsource every problem to AI, your problem-solving muscle weakens. Creativity weakens. Memory weakens. The brain is use-it-or-lose-it. Keep thinking, writing, debugging, and sitting with hard problems on your own. That is where competence lives. 

The AI companies are also in a marketing war right now, and everything is being glorified, maximized, and praised. Some of it is real. A lot of it is sales. Stay grounded.

Every cybersecurity field gets touched. 

GRC (Governance, Risk, and Compliance) people can draft policies faster, map controls to frameworks, and generate testing evidence with assistance. SOC (Security Operations Center) analysts can speed up triage, summarize alerts, and write incident reports in minutes instead of hours. Pentesters can build custom tooling on the fly, summarize recon output, and produce client-ready reports without losing an evening. Defenders get faster. Attackers get faster too.

But AI is also a new attack surface. Prompt injection, model extraction, data poisoning, agent abuse, supply chain attacks on AI tools. These are real, and they are expanding. New standards are showing up to cover them, like the OWASP (Open Worldwide Application Security Project) Top 10 for LLM (Large Language Model) Applications, NIST (National Institute of Standards and Technology) AI RMF (Risk Management Framework), and ISO/IEC 42001 for AI management systems. Whichever side of the fence you are on, red, blue, or compliance, this is now part of the job. 

The professionals who pair security depth with AI fluency will lead this space. The ones who don’t will follow them, or fall behind them.

It comes down to one choice. You either learn to use AI well, or you let someone who does take the seat you wanted.

Beat AI, or let AI beat you.