AI in 2028: From Coding Copilot to Small Business Superpower

AI isn’t just getting a bit better every year—it’s improving so quickly that what feels impossible today could be normal in just a few years. The next 2–3 years are likely to bring a huge shift in how we build software, run companies, and even hire talent.

Below, we break down what this future might look like, why AI agents are such a big deal, and how this all ties into a coming wave of small business creation.

AI Capabilities Are Doubling Every Few Months

Top AI labs expect rapid progress to continue for at least the next two to three years. One way to understand this is through a metric used in research: how long an AI system can work on a task before a human needs to step in and correct it.

A few years ago, coding models could go only a few seconds of “human-equivalent work” before making a mistake. You might get a single correct line of code, then immediately have to fix the next one.

Today, frontier models can effectively handle the equivalent of 10–20 hours of human work before needing intervention. In one benchmark, a leading model reached around 18 hours of continuous, correct work.

Importantly, this doesn’t mean the AI literally spends 18 hours on the task. It usually completes the work much faster. The “hours” measure how long it would take a human to do the same amount of work.

What’s wild is the growth rate: this capability has been roughly doubling multiple times per year. That means every few months, the amount of work an AI can reliably do without help jumps again. As we’ve seen in other areas—like AI video and music—this kind of exponential curve can suddenly make new workflows practical. (If you’re curious how this looks in creative fields, check out how people are using free tools to bulk-generate stylized videos.)

From Co-Pilot to Project Owner: How We’ll Work With AI

As AI gets more capable, the way we interact with it has to change. When models could only handle a few seconds of work, you had to watch them like a hawk—prompt, check, correct, repeat. They were more like autocomplete on steroids.

When models can handle days or even weeks of work, the form factor flips:

• Less hand-holding: Instead of guiding every line of code, you’ll give higher-level goals.

• Bigger, outcome-based tasks: You might say, “Make this app significantly faster, run performance tests, and ensure nothing breaks,” and let the AI run the whole process.

• Event-driven workflows: AI agents will kick off work based on triggers—like a new security vulnerability report or a performance regression—without a human starting every task.

Examples of what this could look like in practice:

• End-to-end performance improvements: An AI agent profiles your app, identifies slow paths, refactors code, runs tests, and opens pull requests.

• Automated security maintenance: It monitors dependency alerts, applies safe upgrades, patches vulnerabilities, and verifies that everything still works.

• Continuous product polish: It reviews analytics, suggests UX tweaks, ships small improvements, and validates them with A/B tests.

This is the world of true AI agents—systems that don’t just answer questions but run ongoing workflows. If you follow AI news, you’ve already seen early signs of this shift in the latest agentic coding tools and frontier models highlighted in roundups like recent AI Weekly coverage.

Why AI Could Trigger an Explosion in Small Businesses

One of the boldest predictions is that AI will massively empower small businesses—and create a lot more of them.

Traditionally, big companies have an edge because they can afford specialization: in-house lawyers, finance teams, senior engineers, product managers, marketers, and more. A solo founder or tiny team has to wear all those hats at once.

AI changes that balance. With strong AI tools, a small team can get:

• Legal gut checks: Quick, high-quality reviews of contracts, policies, and basic compliance questions from AI legal assistants.

• Serious financial analysis: Automated cash flow modeling, scenario planning, and metric tracking without a full-time analyst.

• Top-tier software output: Production-grade code, refactors, tests, and documentation from coding agents.

Instead of needing a large staff, a small business can lean on AI for much of the specialized work and focus human effort on vision, relationships, and decision-making. For people willing to take initiative, this is a huge opportunity: you can realistically build and run something meaningful with far fewer people than before.

Rethinking Hiring: Testing How People Use AI, Not How They Compete With It

As AI becomes a core part of everyday work, the way companies hire is changing too. A growing number of AI-native teams are no longer trying to “block” AI during interviews. Instead, they expect candidates to use it.

In the old model, interviews often tried to measure how good someone was at tasks that AI can now do quite well—like writing boilerplate code by hand. That’s becoming less relevant. The new question is: how well can someone work with AI?

Some teams now run interviews like this:

• AI use is encouraged: Candidates are told they can use as much AI as they want.

• Project-style challenges: Instead of tiny puzzles, they get a few hours to build a real product surface or mini-application—something too big to do manually in that time.

• Evaluation shifts to judgment: Interviewers look at what the candidate chose to build, how they scoped it, what trade-offs they made, and how they used AI to move faster.

The focus moves from raw implementation speed to product sense, problem decomposition, and high-agency execution. In other words: if AI can already write decent code, the differentiator is knowing what to build and how to direct the tools.

Why So Many Former Founders Are Joining AI Teams

Another interesting trend is the number of former founders joining ambitious AI companies instead of starting something new right away. In some teams, entire groups—like “special projects engineers”—are made up almost entirely of ex-founders.

There are a few reasons this makes sense in the current AI moment:

• The problem space is huge: Solving software engineering with AI, or building robust AI agents, is big enough that it can absorb the ambition of many founder-type personalities.

• High-agency work: These roles often blend engineering, product, customer conversations, and commercial impact. You’re not just coding—you’re effectively running mini-startups inside the company.

• Faster learning loops: Being surrounded by other highly driven builders in a fast-moving AI environment can be more attractive than going solo, especially when the tech frontier is shifting every few months.

For companies, this kind of hiring strategy makes sense when the mission is extremely ambitious and the only real constraint is how much ownership and initiative people are willing to take.

How to Prepare for the Next 3 Years of AI

If these predictions play out, the next few years will reward people and teams who adapt quickly. A few practical takeaways:

• Become “agent-forward”: Don’t just use AI as a chatbox—start designing workflows where AI does end-to-end tasks with minimal supervision.

• Think in projects, not prompts: Practice giving AI higher-level goals and evaluating outcomes, not just asking for snippets or quick answers.

• Build small, ambitious: If you’ve ever wanted to start a small business or side project, AI is rapidly removing many of the traditional barriers.

• Update how you evaluate talent: Whether you’re hiring or job-hunting, lean into environments where using AI is expected—and where judgment, product sense, and initiative matter more than manual grind.

The world of AI is changing every few months. By 2028 or 2029, we may look back on today’s tools the way we now look at dial-up internet: impressive for their time, but clearly just the beginning.