The AI Framework That Drives Real Business Results

What actually matters when you’re thinking through how AI fits into your operation, your team, and your career.

What You’ll Learn

• Why the speed of AI change creates a unique adoption crisis most teams aren’t prepared for
• How to audit your existing tools before adding another one (and why this stops most people)
• The real reason knowledge bases matter more than any specific AI tool you buy
• How to set up human-in-the-loop workflows that actually work with the way your brain communicates
• The honest conversation about job displacement and what you should actually be concerned about

Introduction

You already feel it. Things are moving faster. Your inbox fills with new tools every week. People on your team mention AI projects, and you’re not entirely sure what they’re building or whether it’s worth your attention. There’s a nagging feeling that you’re either falling behind or missing something important.

The truth is simpler than it feels. You’re not behind because you don’t have the newest AI tool. You’re behind because nobody’s actually articulated what the strategy should be. And your marketing director AI strategy right now is probably “use ChatGPT when we need copy” or nothing at all.

I’ve spent the last three years watching this play out. Early adopters racing ahead, everyone else frozen by choice paralysis or legitimate concerns about where this is going. The companies that have figured something out aren’t smarter or richer. They’ve just answered a few specific questions first.

This is what I’ve learned about building that framework.

This isn’t theoretical. These insights come from watching teams actually try this.

The Speed Crisis

Every technology advancement has displaced people. That’s not new. What’s different about AI is the speed. There’s no smooth transition window. There’s no learning period where early adoption feels optional.

I noticed this watching Twitter in late 2024. People were already living in 2026. Real-time experiments with new models, tools, and techniques. By the time the same insights made it to LinkedIn, they were ancient history. A month later, they’d trickle into mainstream business publications.

The weird part? Even people working in AI felt behind. The early adopters couldn’t keep up. When that happens, the general population isn’t just behind. They’re spectating a game they don’t understand.

I recognized AI’s potential early enough to bet on it. Same instinct that made me advocate for SEO-optimized websites in 1999, back when agencies treated search engines like a passing fad. Same instinct that pushed me to be the first in my creative department proposing PDFs for remote presentations when everyone else still printed everything. When you see a shift that fundamental, you go all-in out of necessity, not ideology.

But I know plenty of people who saw the same signals and froze. Fear is the actual barrier. Not capability. Not access. Fear of being replaced, fear of breaking something, fear of adopting the wrong tool and looking foolish. So instead of learning how to work alongside AI, they waited. And waiting became falling behind.

The frame shift matters: you’re not at risk of being replaced by AI. You’re at risk of being replaced by someone using AI. You’re not trying to compete with machines. You’re trying to compete with humans who’ve learned how to use them better.

The Data Ethics Friction

Here’s the uncomfortable truth. AI companies trained massive models on creative work, written content, images, and data without asking permission or paying the people who created it. There’s no real accountability mechanism. Your work might have been in that training data. Your client’s might have been.

This created a legitimate outcry in creative communities. Visual artists, writers, illustrators, musicians. The criticism is real. It’s justified.

But here’s the other truth. That legitimate criticism has become a barrier to adoption. I run into people who won’t engage with any AI because they’re morally opposed to the training practices. I get it. I understand the position.

The frustrating part is that staying frozen doesn’t solve the problem. The models exist. They’re being used. Sitting on the sidelines doesn’t change the training practices retroactively. What it does do is leave you further behind while other people figure out how to work with these tools responsibly.

My honest take: the criticism is real, but it can’t be your excuse for inaction. That’s a conversation for policy makers and AI companies. For you, right now, the question is different. How do you participate in this shift while maintaining your own ethics and protecting your team?

Audit Before You Adopt

The most common business mistake I see: teams grab a new AI tool without understanding what they already have.

I’ve watched marketing departments buy tools without checking whether they already have something that does 80% of the same work. They don’t know what tools exist in their stack. They don’t know who’s actually using them. They can’t measure what benefit they’re getting. So they keep buying, layering more tools on top, until the bill becomes absurd and nobody can articulate the ROI.

A real tech audit should happen once or twice a year. Not as a bureaucratic exercise. As a baseline. What tools do we have? Who uses them? For what? What would we lose if we removed it? What problem does it actually solve?

Without that baseline, you have no benchmarks. You can’t measure improvement. You can’t make intelligent decisions about what to add next.

For creative departments specifically, take the AI Opportunity Assessment. Not because I built it. Because the questions force you to think clearly about where you actually are. That’s the first step. Understand your current state before you change anything.

Knowledge Base Is the Foundation

This is the most important part of this whole thesis. It’s also the part most people misunderstand.

Here’s what I see: a company’s documents are scattered across cloud storage, computers, email, old internal wikis, Notion databases, Google Drives. No source of truth. Just chaos that looks organized because it’s hidden in different places.

HR documents are a perfect example. You create a handbook. Five years later, you’ve got thousands of versions floating around. Some are outdated. Some contradict each other. Employees reference whichever one they remember seeing.

AI’s power depends entirely on structured, centralized, accessible information. The better your knowledge base, the better every AI tool you use becomes. It’s not about the fancy model. It’s about what you can actually feed it.

Silos kill effectiveness at scale. Your social media analytics are invisible to the creative team developing next month’s campaign. They’re working blind. Data exists. It’s just locked away.

Video content especially gets short-changed. You’ve got hours of recorded training sessions, customer interviews, case study videos. The only searchable version is whatever auto-generated captions YouTube created, which are often wrong. So that information might as well not exist.

I worked on this problem specifically when building the Creative Brief Builder. I realized the system was only as good as the information feeding it. So the actual innovation was a architecture that let creative teams build a brief-to-launch system using their own knowledge base. Every project became better because the system had access to what the team actually knew.

That’s the inflection point. Not the fanciest AI. A clean, organized, searchable knowledge base. Your existing tools become 5x more useful when they can access good information.

Human-in-the-Loop Is a Communication Problem

This phrase gets thrown around like it’s obvious what it means. It’s not. Most people implementing HITL (human-in-the-loop) workflows think it means “have a human check what the AI did.”

That’s part of it. But the real skill is different.

The process has three steps: set it up, oversee it, review it. But here’s what most people miss. Humans are terrible at giving clear instructions, especially to systems that don’t think like we do.

Think about a game show where one person stands blindfolded at point A and needs to reach point B. Another person verbally directs them. “Turn left. Walk forward. Turn right slightly.” Simple in theory. In practice, misunderstandings happen constantly. “Your left or my left?” “How far is forward?” “Slightly how much?”

That’s human-in-the-loop. It’s an iterative feedback loop. Human tries to explain what they want. AI executes based on what it understood from the explanation. Human sees the output and realizes the miscommunication. Human corrects course. Human tries again with better language. This cycle repeats until the output matches intent.

The real skill isn’t oversight. It’s communication. It’s recognizing that what you say isn’t always what you mean, and that the AI is responding to what you literally said, not what you meant to say.

Once you understand that, HITL becomes a lot more workable. You’re not trying to be a perfectionist overseer. You’re in an iterative conversation with a tool that interprets your language literally.

AI Is a Tool (The At-Bats Theory)

Photoshop didn’t make you a better designer. It made you a faster designer. You could work faster. You could experiment more. You could fail cheaply and iterate.

AI is the same thing. It doesn’t make your thinking better. It lets you work faster.

Here’s the frame that matters: AI gives you more at-bats. If your workflow used to generate 20 ideas, now you can generate 40, 50, or 60 in the same amount of time. Same output from you. More raw material to work with.

This is straight baseball analytics. Your odds of hitting a home run in three at-bats are terrible. Your odds of hitting a home run in 1,000 at-bats are actually pretty good.

The risk is accepting “good enough” as the final answer. Sometimes that’s actually the right call. Shipping a good campaign that’s live beats a perfect campaign that’s still in draft. Speed matters in real business.

But knowing the difference between “good enough to ship” and “actually good” is the human’s job, not the AI’s. The tool gives you more options. You decide which one gets the green light.

The Displacement-to-Rehiring Gap

Let’s talk about the actually scary part. Not the philosophy of AI replacing humans. The economics of it.

AI won’t collapse. It’s too big, too useful, too embedded. Companies will find new niches. The technology will keep improving. That’s not in question.

The real concern is the gap between displacement and new job creation. People are losing jobs right now. Call centers have been slashed. Entry-level creative positions have been cut. Social media agencies are running skeleton crews. That’s happening.

The companies aren’t hiring back. They’re just more efficient. That creates a catch-22. Businesses are more profitable because they need fewer people. But they’re also producing products fewer people can afford to buy. Eventually that math breaks.

This needs serious attention. Not as a nice-to-have. As a business problem that affects everyone’s long-term viability.

I think companies building AI have a responsibility here that most are avoiding. Universal basic income or some equivalent becomes necessary when productivity gains displace more people than new jobs are created. The companies making money from that displacement should be supporting policy that addresses it.

Even Hollywood is feeling this. Entertainment economics are about to shift in a way most people aren’t expecting. The monopoly on what gets produced is already cracking. The studio system worked because production was expensive. AI made it cheap. Now anyone with the right AI tools, the right skills, and a great story can produce high-quality movies. Companies like Higgsfield are already putting out 15-minute episodes that are genuinely entertaining. Within a year, AI video quality will be indistinguishable from traditional production. Netflix should probably be in active talks with AI creators right now, because audiences care about whether something is entertaining, not how it was made.

These aren’t theoretical concerns. These are changing the actual job market right now.

Businesses Need Different People

Here’s what happens when you adopt AI without changing your hiring or organizational structure.

You get more productivity out of the same people doing the same job. That’s good. But it’s not the real opportunity.

What you actually need is more subject matter experts overseeing AI. Domain specialists, not generalists. Your copywriter knowing more about marketing than AI. Your designer knowing more about visual communication than machine learning. That expertise is what keeps the machine honest.

You also probably need an AI CIO role. Someone whose job is managing how AI integrates across the business. Not technical in the coding sense. Critical thinking and communication. Someone who can explain what the AI is actually doing, catch when it’s going off the rails, and translate between what business leaders want and what AI developers think they said.

The other opportunity: productivity gains shouldn’t automatically mean longer hours. You could look at it differently. Same output. Fewer hours. Better work-life balance. Some companies are starting to do this. Most aren’t thinking about it yet.

Those are the structural changes that actually matter.

Security Is the Silent Risk

Here’s what most people don’t know. By default, AI tools use your inputs as training data for the model. Unless you turn it off.

Throw a confidential client brief into ChatGPT without disabling data sharing, and you’ve potentially added that information to the training data. It might be anonymized. That’s small comfort if someone’s clever enough to work backward.

There’s no consistent guardrail across the industry. AI companies are running wild with different privacy policies, different defaults, different levels of transparency. It’s the social media addiction playbook all over again.

The complacency trap is real. AI is right so often that people stop checking. You ask it something, it gives an answer, the answer is accurate, so you assume the context is also right. Accuracy and context are different things. An AI can accurately describe a process while being completely wrong about why that process exists or whether it should exist.

Government regulation will probably eventually matter here. But government moves slowly. By the time there’s meaningful regulation, there’s already billions in usage.

You need to self-protect now. Understand what data you’re sending where. Know which tools are sharing and which aren’t. Have guardrails in place before something sensitive gets leaked.

AGI: When Do We Get There?

Every few months someone declares we’ve achieved AGI (artificial general intelligence). The goalpost keeps moving because there’s no agreed definition.

Jensen Huang, the CEO of NVIDIA, was on the Lex Fridman podcast talking about how they’ve achieved AGI. In context, he meant they could run a $1 billion company using AI. That’s a functional definition of AGI for his purposes.

Without clear benchmarks, there’s no way to know when we’ve actually crossed a threshold. Even experts regularly underestimate what AI can already do. I read analyst notes from 2023 saying GPT-4 wouldn’t be able to handle complex reasoning. In reality, it handles complex reasoning well enough to solve real problems, though imperfectly.

So when will AGI actually arrive? That depends on your definition. If it means AI that can do anything a human can do, we’re probably years away. If it means AI that can do anything a human can do in a specific domain, that’s already happened in chess, image generation, and writing.

The most honest answer is that we’re in the middle of the transition right now. AGI isn’t coming. It’s here, just unevenly distributed.

The Real Insights

Here’s what I believe after more than three years of watching this unfold.

The companies that will win aren’t the ones with access to the fanciest AI tools. They’ll be the ones who audited their existing landscape, built a real knowledge base, put humans in the loop of their workflows, and hired for domain expertise.

The speed crisis is real. But it’s not about speed of technology adoption. It’s about speed of decision making. Companies that can decide quickly, experiment cheaply, and learn from failure will pull ahead. Companies that wait for certainty or perfect information will watch from behind.

The opportunity isn’t bleak. It’s real. But the window to prepare thoughtfully is closing. Every month you wait, the gap widens between where you are and where your competitors are.

The first step isn’t buying something new. It’s understanding where you are. Audit your tools. Understand your knowledge base. Put humans in the loop. Use AI as a tool to do what you already do faster, better, with more iterations.

Then watch what happens.

Key Takeaways

• Audit your existing tools and processes before adding AI; most teams don’t know what they already have or what benefit it provides
• Build a centralized knowledge base first; every AI tool becomes exponentially more useful when it can access organized, searchable information
• Think of AI as expanding your at-bats, not replacing your judgment; the goal is more iterations faster, not automation of decisions
• Human-in-the-loop works when you understand it as iterative communication, not perfect oversight; clarity beats perfectionism
• Speed of decision-making matters more than speed of technology adoption; companies that can experiment, learn, and adjust quickly will pull ahead

Sources

1. Huang, Jensen and Lex Fridman, “Interview: Jensen Huang on AI and the Future of Computing,” Lex Fridman Podcast, 2024, https://podcasts.apple.com/us/podcast/494-jensen-huang-nvidia-the-%244-trillion-company-the/id1434243584?i=10007568063574, Accessed March 31, 2026
2. Acemoglu, Daron and Simon Johnson, “Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity,” PublicAffairs, 2023
3. Sejnowski, Terrence J., “The Deep Learning Revolution,” MIT Press, 2018 https://mitpress.mit.edu/9780262038034/the-deep-learning-revolution/,