The Agentic Age Is Here: What Changes for a Founder of One

The AI age is already over.

It lasted about three years. Then something shifted, and it wasn’t a new model or a breakthrough paper. It was a recognition: AI isn’t a tool anymore. It’s infrastructure.

When a technology becomes infrastructure, we stop naming businesses after it. Marc Andreessen wrote in 2011 that software was eating the world, and within a few years nobody said “internet business” or “mobile business” anymore, because everything was. The same thing is happening to “AI business” right now.

What comes after the AI age is the Agentic Age.

If you’re building a business today without understanding that transition, you’re building something that will feel old before it launches. I know because I’ve done it.

What “agentic” actually means

Precision matters here, because the word is being stretched to cover everything.

AI: a system that performs a task requiring something like intelligence: recognizing a pattern, processing language, making a judgment.

Agentic AI: a system that doesn’t just perform a task when prompted, but works out what needs doing next and does it, across many steps, using tools.

Anthropic’s engineers drew the line cleanly in Building effective agents (December 2024): workflows are “systems where LLMs and tools are orchestrated through predefined code paths,” while agents are “systems where LLMs dynamically direct their own processes and tool usage.” Most of what runs in my studio is the first kind. The interesting part is how much of the work between my decisions the second kind can now carry.

The difference in practice:

• AI answers a question. Agentic AI decides which questions to ask.
• AI follows an instruction. Agentic AI works out the instructions from the goal.
• AI helps you do the work. Agentic AI does the work between your decisions.

Think of it this way:

• AI-age thinking: “How can I use AI to answer support emails faster?”
• Agentic-age thinking: “How do I build a support system that triages, answers what it can, and escalates only what needs me?”

The first adds AI to an existing process. The second redesigns the function around agents.

Why now

Three things converged.

1. The models got good enough to trust with routine judgment

Early language models were impressive demos and unreliable operators. They’d hallucinate a citation, break a build, or confidently do the wrong thing.

For me the threshold wasn’t a model number. It was the week I stopped reviewing every line an agent wrote and started reviewing every pull request instead. That’s a different job. Line-by-line review means the agent is a typist. PR review means the agent is a colleague whose work I check before it ships.

Sufficiency, not perfection, is what matters. You don’t need an agent that’s smarter than you on your best day. You need one that’s more reliable than you on your worst day, at two in the morning, on the tenth repetition of a task you’re bored of.

2. Orchestration became possible

A single model call is a component. The shift came when you could compose agents: one researches, one drafts, one checks the draft against a spec, one decides whether a human needs to look.

No single agent is magic. The system is what matters.

That’s what the Fractary Platform is for. Faber carries a piece of work from an issue to a pull request through five phases, with checks between them: agents do the work inside each step, and code runs the loop around them. (What that looks like at scale is its own post.) Codex keeps the documentation those agents rely on in sync across every repo. The plumbing between agents and tools is increasingly standard too; the Model Context Protocol, which Anthropic released in late 2024, is the reason a tool I wire up once works across the agent runtimes I use.

3. Economic pressure

For a company, the pressure is headcount cost and capital that stopped being free.

For a founder working alone, the pressure is simpler and more brutal: I have one person’s hours. Every hour I spend on routine work is an hour a venture didn’t get. There’s no budget line to add people to. Either agents take the routine work, or the routine work takes the venture.

That’s the pressure that turned agentic architecture from interesting to mandatory for me.

AI-added vs. AI-native

Here’s the fault line.

AI-added businesses take existing processes and bolt AI on:

• The support inbox, plus a chatbot
• The sales process, plus an email assistant
• The weekly report, plus an AI summary
• The product, plus an “AI-powered” feature

AI-native businesses design the function around agents from the start:

• Support that triages and resolves, and escalates the rest with full context
• Outreach that researches and drafts, and hands you the conversation
• Analysis that runs continuously instead of on a calendar
• Products that adapt without a release cycle

The AI-added approach gets you a real but modest improvement. The AI-native approach changes what one person can run. I won’t put a multiplier on that, because every number I’ve seen attached to it is either a vendor’s or a guess. The honest claim is that I couldn’t run a dozen ventures the other way, and I’ve tried.

What changes for a founder of one

If you accept the premise, here’s what actually changes.

1. Structure

Old model: functions organized by department, executed by people.

Agentic model: systems organized by outcome. A revenue system. A customer-success system. A build-and-ship system. Humans design and monitor; agents execute.

If you’re alone, this is the only structure that exists. There are no departments. There are the things you do by hand and the things that run.

2. The hire you don’t make

Old question: “What role do I need to hire for?”

Agentic question: “What in this role requires human judgment, and what’s a system?”

For a solo founder, the answer to “should I hire?” is usually “not yet, but I should build the system that a hire would have run.” When I catch myself wanting to hire, that’s the signal to write down what the person would do. Most of it turns out to be a workflow.

3. Capital

Old math: revenue scales with team size and burn.

Agentic math: revenue scales with how much of the operation runs without you.

For a founder of one, this means the business can be profitable at a size that would never justify a team. That opens up ventures that were never economically viable before, which is most of what I build.

4. Moats

Old moats: brand, network effects, data, distribution.

Agentic moats: how fast your systems improve compared to the competition’s. The traditional moats still matter. They’re just table stakes now.

5. Where the hours go

I won’t give you a percentage breakdown of my calendar, because I’d be making it up. What I can tell you is the direction: less time executing, more time designing the system that executes, and more time on the two things agents are worst at, which are talking to customers and deciding what not to build.

The thesis in action

I’m not theorizing. I run a dozen ventures alone, and every one of them is AI-native from inception, because the alternative doesn’t exist for one person.

Most of them are early, and the ventures page says exactly how early. A few examples:

• Corthos, in private beta, is a suite of technical tools and infrastructure for powering autonomous organizations: engines for acquiring, transforming, storing, and publishing data.
• VoxBuy, a prototype, is a product-decision tool that maps options on a value graph to find the honest best picks, with no paid placements.
• Org Strong, in preview, is AI tooling for volunteer-run community organizations, so that institutional knowledge doesn’t leave when the volunteer does.

None of these would exist if I had to staff them. All of them are possible because the routine work is done by systems I designed once.

The strategic inflection point

Andy Grove defined a strategic inflection point in Only the Paranoid Survive (1996) as “a time in the life of a business when its fundamentals are about to change.”

We’re at one.

The businesses being built today will split into two groups:

1. Legacy-compatible: built on the assumption that humans execute and AI assists.
2. Agentic-native: built on the assumption that agents execute and humans architect.

Both can succeed in the short term. My opinion, and it’s an opinion, is that the economics of the second group are strong enough that the first group becomes hard to defend within a decade.

This isn’t technology for its own sake. The mechanics are plain:

• Lower operating cost → higher margins → pricing flexibility
• Faster execution cycles → faster learning → better fit
• Systems that run around the clock → better customer experience
• Growth without proportional headcount → unit economics that work at small scale

What this means for you

If you’re building something right now, you have three options.

AI-added. Keep the conventional model and add AI features. Easier now. Safer by the old metrics. You’ll compete against agentic-native businesses with structurally better economics.

AI-native. Design around agents from the start. Harder now. Requires skills you may not have yet. Hard to explain to anyone using the old playbook. Compounding advantages later.

Hybrid. Start AI-added but architect for the transition. Realistic for an existing business. The risk is getting stuck in the middle.

If you’re starting from zero, and especially if you’re starting alone, the choice is easier than it looks. You don’t have a legacy to protect. Build it the new way.

The counterintuitive part

Here’s what most people get wrong about the agentic age:

It isn’t about AI getting smarter. It’s about humans getting better at designing systems that agents can run.

The technology is already good enough for most of what a small business does. What’s missing is the design thinking: the habit of asking “what would this look like if I couldn’t do it by hand?” before building anything.

That’s why The Fractary is a studio and not a single product. I’m not betting on one application of the pattern. I’m testing whether it holds across very different kinds of business:

• Commerce (VoxBuy)
• Community organizations (Org Strong)
• Education (eCollective: DIY Degree, Open Degree, InstructOS)
• Healthcare and data intelligence (Locutos, Infactas)
• Digital identity (OmniDAS)
• Infrastructure (Corthos, the Fractary Platform)

Different verticals, one set of principles. If the pattern holds across that range, it isn’t a niche trick.

The question that matters

Forget whether AI will replace jobs, or whether regulation will slow it down. Those are interesting but not actionable this week.

The question that matters for your business is:

If you were building it from scratch today, knowing what agents can do, would you build it the way it runs now?

If the answer is no, you have work to do.

If the answer is yes, you’re either already agentic-native or you haven’t yet seen what’s possible.

The playbook

Based on building a dozen ventures this way, most of them still early, here’s what I’d tell anyone starting:

1. Start with outcomes, not roles

Don’t ask what positions you need. Ask what outcomes you need and the minimum human involvement to get them.

2. Default to a system

Every time you want to hire, or want to do it yourself again, first design the system that could handle most of it. Keep for yourself the part that truly needs judgment.

3. Build the intelligence into the plumbing

Monitoring, checking, and summarizing should happen without anyone asking.

4. Design for absence

Build systems that keep working when you’re not there. I wrote about this in Freedom Through Autonomy.

5. Measure the system, not your busyness

Track whether outcomes happened, not how many hours you spent.

The future is unevenly distributed

William Gibson’s line, “the future is already here, it’s just not evenly distributed,” is the best description of this moment I know.

The agentic age isn’t coming. It’s here. It’s just concentrated in a small number of builders who noticed early and rebuilt their foundations.

I’m one of them, with a dozen early ventures and a lot still to prove. There are others. More every month.

The question isn’t whether this transformation happens. It’s whether you’re early enough to benefit from it.

Your move

Some businesses being built right now are AI-added. They’ll improve incrementally on last century’s model.

Others are being built agentic-native from day one. They’ll create categories that weren’t economically viable before, including the category I care most about: real businesses run by one person.

Which are you building?

If you’re a teenager, or you have one, the fastest way I know to find out what one person plus agents can do is to build a first business that way. That’s what Make First is.

---

This is the second post in a series on building for the agentic age. Previous: Freedom Through Autonomy: Businesses That Run Themselves. Next: Building the Fractary Platform: Why I Open-Sourced My Stack.