The Future of AI for Business Isn't About Replacement

Key Points

The AI wave is splitting businesses into two camps: those leveraging it for efficiency gains and those still experimenting with no clear ROI
The real value isn’t automation replacing people—it’s augmentation that lets smaller teams punch above their weight
The winners in the next 18 months won’t be the ones with the biggest AI budgets; they’ll be the ones with the clearest problem statements

We’re past the hype phase now. It’s April 2025, and the AI conversation in business has shifted dramatically. A year ago, everyone wanted to “add AI” to their product. Today, the smarter companies are asking a harder question: “Which specific problems does AI actually solve better, and how do we measure that?”

I’ve been using AI tools daily at Rotate since Claude 2 launched. I’ve tried probably two dozen different generative AI applications across my businesses. And I’ve watched clients wrestle with the decision to integrate AI into their operations. What I’m seeing isn’t the exponential growth curve from the demos—it’s something messier and more real.

The Implementation Gap Is Real

Here’s what I’ve noticed: there’s a massive gap between “AI can theoretically do this” and “we’ve actually implemented this in a way that moves the needle.” A lot of companies built elaborate AI initiatives in 2024 that are quietly being shelved in 2025. Not because the technology doesn’t work, but because the problem wasn’t actually there to begin with.

I worked with a client last year who wanted to replace their customer support team with an AI chatbot. They had seven support agents. The math seemed compelling—automated support would cut their costs in half. Except when we actually looked at what their support team did, most of their issues were complex customer onboarding problems that required human judgment, and a few were straightforward questions that got asked once per quarter. The expensive team was actually the right answer.

That’s become the real pattern I’m seeing: companies confuse “could we use AI for this” with “should we use AI for this.” The ones making real money with AI right now are asking the second question first.

Where the Real ROI Is Happening

The winners I’m watching fall into a few clear categories.

Content-generating businesses are having a legitimate moment. We’re helping clients use AI to scale their content output—not replace writers, but give them a force multiplier. A writer who uses Claude or ChatGPT for research, outlining, and editing can produce maybe three times as much work in the same hours. That’s real value. Ryan’s Roundup, my newsletter with 12,500 subscribers, couldn’t exist without AI tools helping me process information faster. I’m not replacing my thinking; I’m outsourcing the grunt work.

Coding is genuinely transformed. I don’t say this lightly—I’ve been writing code for fifteen years. Cursor with Claude and GitHub Copilot have changed my velocity as a developer. Not by replacing me, but by handling the 30% of coding that’s boilerplate and lookup work. The 70% that requires judgment, architecture decisions, and domain knowledge—that’s still all me. But I’m shipping features two or three times faster because I’m not reading documentation or building the fifteenth form validation function. At Rotate, we’ve seen client projects compress timelines by 20-30% with the right tooling.

Internal operations are where I’m seeing the most reliable gains. Scheduling, basic data analysis, documentation, email drafting, meeting notes—the work that doesn’t directly generate revenue but definitely slows things down. One of my businesses uses Claude to process customer feedback and categorize it. Used to take someone 4 hours a week. Now it’s 30 minutes for a human to review Claude’s categorization. That’s not sexy, but it’s money.

Research and analysis for decision-making. This is where I use AI the most in my own work. Instead of spending three hours reading, I can have Claude help me synthesize information, find patterns, and push back on my assumptions. For someone like me who runs multiple businesses, that’s invaluable time back.

What’s NOT working at scale? Wholesale replacement of human roles. Every company I know that tried to automate their way out of payroll costs ended up rebuilding those positions in different forms because the nuance got lost.

The Solopreneur Advantage Is Real

Here’s something I’m genuinely excited about: if you’re a solopreneur or you run a small team, 2025 is your moment. AI doesn’t level the playing field between you and big corporations—it tilts it.

A solo consultant can now do research, create content, manage their operations, and handle customer communication with a fraction of the support overhead that would require hiring someone five years ago. I know people running six-figure businesses solo now with Claude and Cursor doing the equivalent work of two full-time employees. The economics are fundamentally different.

The catch is that you have to actually use the tools skillfully. Prompt engineering isn’t a meme—it’s the difference between AI giving you garbage and AI giving you something worth refining. The solopreneurs winning are the ones who’ve spent time learning how to get good outputs, not the ones who tried it once and said “AI can’t write” or “AI can’t code.”

What’s Overhyped and What’s Real

Let me be direct about what hasn’t delivered:

AI agents doing complex workflows end-to-end. They sound amazing in demos. In practice, they’re fragile. They work great on controlled problem sets and fall apart in the real world. I think agents will be important in 18-24 months, but right now, if you’re betting your business on them, you’re betting on something that’s still in beta.

Domain-specific AI tools that promise to run your business. Most are just ChatGPT wrappers with nicer UI. Some have real value (industry-specific training data, integration with your systems), but a lot are solving problems that a good prompt in Claude already solves. Be skeptical.

Replacing your knowledge workers outright. This is the narrative that gets media attention and drives fear, but I’m not seeing it in practice. What I’m seeing is knowledge workers who use AI tools outcompeting knowledge workers who don’t. The people losing jobs aren’t losing them to AI—they’re losing them to people who use AI better.

What’s actually real:

Claude and GPT-4o are legitimately useful. I’m not being paid by Anthropic, I just use both daily and they’re genuinely different in useful ways. Claude is better at reasoning and code. GPT-4o is better at image understanding and has a faster API. Both are tools worth having.

Code generation is a game-changer for developer productivity. This one’s done. It’s shipped. Cursor is a brilliant implementation of it. If you’re a developer not using an AI coding assistant in 2025, you’re choosing to work slower than your peers.

LLMs are genuinely good at analysis, synthesis, and pattern-finding. Not perfect, but useful. If you’re processing information, they can save you real time.

What I Think Happens in the Next 18 Months

By late 2026, I think we’ll see a clearer separation between companies using AI strategically and companies that tried it, didn’t see immediate ROI, and moved on. The experimental phase is ending.

I expect more vertical AI tools with real domain expertise—not just wrappers. The accounting software that uses AI to categorize expenses and find anomalies will beat a spreadsheet plus ChatGPT. The CRM that uses AI to write follow-up emails in your voice will beat copy-pasting prompts. That’s valuable integrations, not general-purpose chatbots.

I also expect the “AI replacing jobs” narrative to cool down as companies realize the ROI story is more subtle: AI augments some roles, eliminates some repetitive tasks, and sometimes creates entirely new roles to manage the AI tools themselves.

Coding will continue to accelerate. The distinction between junior and senior developers won’t disappear, but it’ll shift. A great developer with AI tooling will do 5x the work of a developer without it.

The solopreneur/small team era will keep expanding. Every year, it becomes more viable to run something meaningful with a tiny team plus AI augmentation. That’s a genuine structural shift in how businesses can operate.

What This Means for Your Business Right Now

If you’re deciding whether to invest time or money in AI for your business in 2025, here’s my honest take: start small and specific.

Pick one workflow that’s currently painful, time-consuming, and well-defined. Spend a week genuinely experimenting with AI tools for it. Not Googling “should I use AI,” but actually using Claude or ChatGPT or an industry-specific tool. Measure whether it actually saves time or money. If it does, expand it. If it doesn’t, move on.

Don’t let FOMO drive you to massive AI initiatives. Don’t ignore it either. The businesses thriving right now are treating AI like any other tool: useful for specific problems, overhyped overall, and worth learning to use well.

We’re still early. But “early” now means “past the hype, into pragmatism,” not “nobody knows what this is yet.” That’s actually a better time to build.

The future of AI for business isn’t about perfect automation or massive disruption. It’s about who figures out how to make the technology genuinely useful for their specific situation. The winners will be the ones who ask better questions, not the ones with bigger AI budgets.