Murph's Take

AI Hype vs. Reality in 2025: What's Real, What's Not

The AI headlines are breathless. The reality for small businesses is more nuanced. Here's my honest take on what's actually working and what's still vaporware.

MurphJanuary 20, 20257 min read

Every week there's another headline. "AI will replace X profession." "New AI model achieves human-level Y." "AI agents will run entire businesses by Z date."

Some of it is true. Some is years away. Some is marketing.

Let me give you a ground-level view from someone who actually helps small businesses implement this stuff.

What's Real Right Now

AI as a Writing Accelerator

This is real, proven, and working today. Not future state. Now.

Any business that produces written content — emails, proposals, website copy, social posts, documentation, follow-up messages — can be meaningfully more productive with AI. Not 10% more productive. 50-70% for the writing portion of work.

I've watched business owners who spent six hours drafting a complex proposal do it in 90 minutes with AI assistance. That's not hype. That's what's happening in real businesses right now.

AI Automation for Repetitive Workflows

Real. Working. Delivering ROI today.

Lead capture, follow-up sequences, meeting summaries, data routing between systems, review requests, appointment reminders — all of this can be automated with current tools for reasonable cost.

The businesses that have done this work are running more efficiently than those that haven't. The gap is real and it's growing.

AI for Research and Synthesis

Real, with caveats.

Tools like Perplexity for current web research, Claude and ChatGPT for reasoning through complex information — these are genuinely useful. They accelerate research. They synthesize large volumes of content. They surface relevant information faster than manual search.

The caveat: they still hallucinate. They still make confident-sounding claims that aren't true. You need to verify anything important. Not a reason to avoid them — a reason to use them intelligently.

What's Partially Real

AI Customer Service

Chatbots and AI customer service are real. Whether they're good is variable.

Basic question-answering, lead qualification, appointment booking — AI handles these well. Nuanced complaint resolution, complex emotional situations, anything requiring real empathy — AI still struggles.

The best implementations I've seen use AI for triage and simple questions, and route complex situations to humans. Pure AI customer service, with no human fallback, creates customer experience problems that often cost more than the savings.

AI "Agents" Doing Multi-Step Work

Partially real. Getting better fast.

AI agents that autonomously research, decide, and act are genuinely improving. I've built workflows that do meaningful multi-step automation. But they require careful design, and they fail in interesting ways when they encounter edge cases.

For business use today: supervised AI agents for specific, bounded tasks. Fully autonomous AI agents handling open-ended business problems? Not ready for prime time.

AI-Generated Video and Imagery

Real in capability, still awkward in execution for most business uses.

AI video is getting impressive fast. But the uncanny valley is still there for anything showing realistic humans. AI-generated images for web use and marketing materials? Genuinely useful. AI video for your client-facing communication? Mostly not yet.

What's Hype

"AI Will Replace Your Entire Team"

No. Not now, not soon.

AI augments humans. It replaces specific tasks within human roles. It may change the headcount requirements for certain functions over time. It will not replace the need for humans in the next 2-3 years in most small business contexts.

Anyone telling you to fire your team because AI can do everything is wrong, or selling you something.

"Autonomous AI Business Operators"

I see tools marketing themselves as "let AI run your business while you sleep."

There are legitimate automation tools that handle specific workflows. There is no AI system today that runs a real business autonomously. Building that is a decade-plus away. Anyone who says otherwise is misrepresenting current capability.

"AI Trained on Your Business That Knows Everything"

The pitch: upload all your documents, emails, and data. AI becomes an expert on your business and answers any question.

The reality: this kind of "AI on your data" is real but much more limited than advertised. Current tools work for specific, bounded knowledge bases. They struggle with ambiguity, conflicting information, and nuanced judgment. They're useful for specific applications — not a general-purpose business intelligence system.

"No-Code AI Anyone Can Build in Minutes"

The tools have gotten easier. They haven't gotten effortless.

Getting real, reliable automation working for a specific business problem still takes time, iteration, and judgment. "Build a full AI stack in an afternoon with no experience" marketing is not matching reality for most business owners.

The Honest State of Play

AI is genuinely transformative for specific tasks. Right now. Real ROI. Not future state.

AI is not yet an autonomous business operator, a full customer service replacement, or a general-purpose expert on your company.

The businesses winning with AI are the ones who've identified specific, real problems, used AI tools to solve them, measured the results, and kept iterating.

The businesses getting burned are the ones chasing the hype — spending on ambitious AI projects that aren't ready yet while missing the unglamorous wins that are working today.

Know the difference. Capture the wins. Build toward the future with clear eyes.

That's the play.

Want this for your business?

Tell us what you're building. We'll map out exactly what to build and what it costs.

Start Your Project →

Frequently Asked

What AI capabilities are actually working for small businesses right now?

Writing acceleration (proposals, emails, content) is real and proven — most business owners see 50-70% time reduction on written work. Automation for repetitive workflows (lead response, follow-up sequences, review requests) is delivering measurable ROI. These aren't future-state — they're working in real businesses today.

What AI promises are still mostly hype for small businesses?

Fully autonomous AI agents that run entire business functions without oversight, AI-generated video at broadcast quality, and voice AI that can't be distinguished from a human in complex customer service scenarios are all still early or unreliable for production use. The gap between demo quality and production reliability is where most small business implementations fail.

How should a small business owner evaluate whether an AI tool is worth the investment?

Ask one question: can you measure the specific outcome it's supposed to produce before and after? If an AI tool is supposed to save time, measure the time. If it's supposed to improve lead conversion, measure the conversion rate. Any tool without a measurable outcome attached to a specific workflow is a speculative purchase.

Is now actually a good time to adopt AI, or is it better to wait for the technology to mature?

The businesses building AI workflows now are accumulating operational advantages — better content libraries, faster lead response systems, more automated operations — that compound over time. Waiting for 'mature technology' has the same structure as waiting to build a website until web design was 'figured out.' The businesses that waited lost years of compounding advantage.

Jason Murphy

Written by

Murph

Jason Matthew Murphy. Twenty years building digital systems for businesses. Former CardinalCommerce (acquired by Visa). Now running VibeTokens — AI-built websites and content for small businesses.

The window is open.

It won't be forever.

Start Your Project →