The AI Agent Surge: How to Make Yourself Irreplaceable Before It’s Too Late

Why the agentic AI revolution will transform work faster than leaders can adapt.

Your competitors are showing up with laser tanks while you're still perfecting your sword technique.

After weeks of procrastination, I'm putting my thoughts about AI agents into the world instead of just letting them bounce around my head at 3AM. I’ve also been mainlining Billie Eilish's latest album all week—that has absolutely nothing to do with AI agents, but something about that blend of hard and soft is the perfect soundtrack for processing how radically our work is about to change.

If you find yourself nodding along, hit subscribe now—this is just the beginning of conversations we need to have about a future none of us are prepared for.

---

Everyone's talking about AI agents, but barely anyone knows what they are.

At least that's what the Wall Street Journal pointed out this morning in their article about the buzzword "agentic" taking over Silicon Valley. And they're right—most businesses lack a common understanding of what these agents actually are or what they'll do to our workplaces.

Look, I've been tracking this for about 10 months now, and at first, the timeline for this technology seemed like pure science fiction. But as we push into mid-2025, the pace is absolutely relentless. I still lose sleep over it.

What are AI agents, really?

Let's cut through the noise. According to the WSJ, AI agents are "systems that can take some action on behalf of humans"—like buying groceries or making restaurant reservations. But that simple definition misses the bigger picture.

Here's the thing: AI agents go beyond LLM systems like ChatGPT because they take unprompted action. They reason and act on your behalf without you having to hand-hold them through each step. They're not just glorified chatbots (which have been around forever). If you have to prompt the AI to do a specific action, it's not an agent—it's just another LLM.

Simply put, AI agents will evolve into digital colleagues. This is a paradigm shift in how we operate in the workforce.

Why the timeline should keep you up at night

I'd encourage anyone reading this to check out Leopold Aschenbrenner’s paper on situational awareness for AGI. He worked on the super-alignment team at OpenAI that managed the risks of superintelligent AI, and according to his paper, we are on course for AGI (artificial general intelligence) by 2027.

It's currently 2025. So two years from now, these AI systems will basically be able to automate all cognitive jobs. Think about that—all jobs that can be done remotely.

Right now, we're using glorified spell-checks and chatbot assistants. The more cutting-edge among us might be playing with image models like Midjourney or ChatGPT's new image generator (which is blowing up this week with all those Studio Ghibli style recreations).

It's easy to get drunk on these tools. If you're a creative person, it feels like magic. But these tools are primitive compared to what's coming next.

Where we're going with AI agents is drop-in remote workers that break from the bounds of their compartmentalized sandboxes and take control of your device. That, to me, is what defines an agent. You can onboard an agent like a human hire.

The agent will message you on Slack. It will take on projects. It will write emails, execute digital task work, and fulfill creative strategies on your behalf. Right now, you are the prompter. Soon, you'll direct the prompter.

This isn't future tech—it's happening now

The adoption is happening faster than you think. Deloitte predicts that 25% of companies using generative AI will launch agentic AI pilots this year, growing to 50% by 2027. We'll see some companies adopt these tools in the back half of 2025.

OpenAI has already released a type of agent called "Operator" that acts as a personal web assistant—sifting through content, ordering groceries, creating memes, booking appointments. This exists today.

If you code, you've probably been exposed to Claude Code. It's the first agentic AI that caught me completely off guard with its capability. Once I tried it, I burned through hundreds of dollars in API credits building powerful MVPs just to test its limits.

You give it access to your entire codebase and prompt it with things like "refactor this entire codebase into TypeScript" or "implement my static code so it works with Contentful's API." You can even run git commands in human language: "commit and deploy this code." It requires minimal prompting once you've told it what you want.

I understand the technicalities of what I'm trying to do, and I lose sleep over its capabilities and what it means for the future. Imagine how powerful that will be when these agent workers break out of their self-contained boxes. Agents will have situational awareness and understanding of our needs in ways we can't even fathom yet.

My instinct is that true computer agent capability will align to Leopold Aschenbrenner’s timeline—within the next two years—but we won't adapt quickly enough because we're distrustful and apprehensive about new technology.

Money talks, though, and nothing will hold back these floodgates. Once enough businesses start implementing agents, it will be impossible not to compete.

What gets upended first (spoiler: it's management)

So far, we've seen these tools implemented for creative use (Midjourney, ChatGPT, Sora) and now disrupting development environments. These were things we couldn't fathom being automated a few years ago.

For its next trick—and this will blow minds—it will come for management tasks. Agent workers will take over low-level HR decisions, scheduling, planning, coordinating meetings, booking events, reviewing budgets, running recurring task work, and preparing reports. I have no doubt this type of work will be driven by agents without prompting.

Remote-First vs. Office-First: Clear winners emerging

Remote-first companies will adapt quicker due to the nature of their digital work environment. Teams who are remote are already primed to work alongside agent tools.

Take note-taking bots that sit alongside video calls and meetings—standard for remote workers but less common for onsite teams. Remote companies will be the first to adopt these agents. Onsite companies should be paying attention.

Real-world impacts you can't ignore.

I see a real benefit in employing AI agents to clean up the technical debt that teams have accrued—tasks we know need to be done but don't have the time or willingness to prioritize.

This includes documenting code bases, archiving projects, organizing Google Drive file systems, organizing proposals and contracts, developing templates, documenting standard operating procedures, closing unused chat channels, and clearing out accounts for pay-per-seat software you might be overpaying for.

These tasks could take humans days to complete, but remote agents could run them automatically as part of a daily routine. If an agent can take initiative and clear out all those straggling bugs in your codebase that you've been parking for a year, that's going to reduce significant cognitive load and anxiety for your team.

The irony is that agents might also make great partners in identifying the most high-value tasks that humans can take on. If it has any self-awareness at all, demonstrating where a team thrives and developing routines that improve the business would be immensely valuable.

I'm optimistic. The opportunity isn't to replace jobs but to reduce the busy task work nobody wants to do. Yes, big companies will cut their workforce, but they will always argue that there was already fat to trim before these tools arrived.

The trade-off is that teams need to be more strategic. People who lack agency, creativity, and curiosity will have trouble adapting to the rapidly changing workforce. If you're a task worker, you need to be aware that these tools are coming whether you like it or not.

How agents will transform team dynamics

I think AI agents will be great for team alignment. They'll excel at identifying where a company isn't aligned on tasks versus the greater goals and values of the organization. They'll be good at scouring data, notifying of shortfalls on targets, and emphasizing corrective action.

To some degree, they'll replace hard conversations that humans don't like to have. That's a boundary we need to be careful about—it will be irresistible for managers to let agents do the dirty work, but we need to maintain our humanity, empathy, and candid nature. That's our advantage as humans.

The other drawback is accountability. When something inevitably goes wrong, who's responsible for a failure in the codebase, a misaligned client communication, or a major mistake in the numbers if the AI agent did all the work? Better checks and balances will need to be implemented, and human ownership will be an expectation in the future workplace.

This isn't science fiction.

These are preliminary thoughts, but none of this is fantasy. This is real. Within the next two years, you'll see much of this come to fruition in your workplace. It's better to understand it now than get caught off guard.

At the end of the day, the battlefield has changed. Your competitors are showing up with laser tanks while you're still perfecting your sword technique. The question isn't whether you'll adopt these tools—it's how quickly you'll adapt when everyone else already has.

---

If you found this valuable, I'd love for you to support my little newsletter. Frontier Notes explores the intersection of technology, leadership, and the future of work without the usual hype or doom-spiraling. Just practical insights from someone in the trenches.

---

P.S. If today’s insight shaved even five minutes off your mental load, please do me one quick favour—forward this note to a teammate who’s staring down the same AI storm. Every share gets us closer to 1,500 sharp-minded subscribers.