You are doing AI wrong + bye Medium

10 Insights on Navigating the AI Revolution

6 min readMar 21, 2025

***NOTE: I am moving to Substack! ***
This will be my last publication here. If you are interested in keep reading what I write, please subscribe here: https://jonathankahan.substack.com/p/you-are-doing-ai-wrong

For organization to successfully adopt AI, many factors need to be aligned. Like a bank vault where every tumbler must click into place before the door swings open. And around us, we are seeing organizations jamming the mechanism all the time because they discounted one factor or the other.

Encouraging people to play with AI to improve the way they work makes sense, firing employees right and left because AI is come is a terrible idea. Communicating excitement about the tech helps, excessive hype hurts. Assuming all companies will be AI-powered tomorrow when many are still struggling to adopt cloud, is deluded. At the same time, we can already see AI changing the way citizens and consumers relate to content, news, and each other, with AI even starting to have an impact on our cognitive abilities.
The hype is real: the signal-to-noise ratio in AI discourse is appalling, and we all suffer from this confusion.

After dozens of conversations with clients, friends and experts, we’ve distilled ten truths about AI that cut through the noise. Spoiler alert: as it often happens, the most underrated factor is the human one.

1. AI is not competitive advantage.

The AI gold rush has a fatal flaw: it’s available to everyone. AI is already commoditizing itself: as LLMs become cheaper and open source, merely implementing AI will no longer differentiate you. Traditional advantages like expertise, processes, and even economies of scale will collapse as AI optimizes everything to near-identical perfection.

What will matter? Access to customers, unique frameworks, memorable brands, and niche, hard-to-collect data. These will become increasingly valuable as they enable truly differentiated AI outputs in a sea of algorithmic sameness.

2. AGI is already here, but it might take a while to notice.

While engineers debate whether AGI exists, organizations fail to use the AI they already have. Even in companies that have invested millions, actual use remains limited to a few enthusiasts. Is that because AI isn’t smart enough? No — most tasks don’t require PhD-level intelligence.

The barriers aren’t technological but human. The two biggest challenges standing in the way of AI adoption are 1. Having effective interfaces to allow task augmentation/automation, and 2. Integrating AI-augmented workflows within an organization’s broader processes. Even if AI advancement stopped today, companies might need a decade to solve these challenges and absorb existing capabilities. Those who solve these integration challenges will extract value where others see only potential.

3. You don’t need an AI tool.

The AI gold rush has everyone buying shovels, but few have a map of the territory.
We are seeing companies rushing to adopt AI tools without clear strategy. It is like buying a Ferrari without knowing how to drive : expensive and potentially dangerous. AI won’t solve lack of strategic focus. It won’t magically establish new processes and it won’t by itself break down silos.
And then of course, effective AI implementation requires a strategy of its own before obsessing over specific tools. The most successful organizations will be those that align AI capabilities with core business objectives rather than chasing technology for its own sake.

4. Automation is just the peak of the iceberg.

Companies buy AI to cut costs when they should be using it to grow in ways they haven’t imagined. AI adoption follows three horizons: First, businesses reduce costs by automating tasks. Second, they leverage AI for creative work, developing better, more customized products. Third, they restructure entire processes around AI capabilities, unlocking completely new value propositions.
Most businesses remain fixated on cost-cutting, missing AI’s transformative potential.

5. Trust is the Ultimate Competitive Advantage

People won’t use AI they don’t trust, no matter how powerful it is. The winners of the AI revolution won’t be those with the most advanced technology, but those who master the human-AI trust relationship.

Hallucination isn’t a bug — it’s all AI does. Every output is a prediction, not a retrieval. Many hallucinations still produce useful results, but confident-sounding errors destroy trust instantly. We need AIs that communicate uncertainty as fluently as they communicate answers. This becomes critical as invisible agents operate beneath interfaces, making decisions we can’t monitor.

When you can’t see what an AI is doing, you need absolute certainty it knows when it doesn’t know. Organizations that solve this confidence-signaling problem will outperform those building ever-larger models that simply hallucinate more convincingly.

6. The Attention Economy’s Final Form

Tomorrow’s algorithms won’t just compete for your attention — they’ll battle for your reality. We’ve spent millennia creating content, centuries making it accessible, and decades trying to filter its overabundance. But this is nothing compared to what’s coming. When AI can generate billions of hyper-personalized rabbit holes, each calibrated to captivate with uncanny precision, truth becomes negotiable and cognitive bandwidth becomes our scarcest resource.
Verification services, trusted curators, and shared authentic experiences will command premium value as we navigate a sea of algorithmically-perfect manipulation. This weaponization of abundance will force radical reinvention of how we structure our daily information diet to preserve independent thought.

7. People skills is the new STEM.

As transactional aspects of work become automated, the relational aspects will gain extraordinary value. The elements that can’t be automated — making people comfortable, coaching, guiding, creating psychological safety — will become central to both business success and human fulfillment. Organizations that recognize and nurture these relational competencies will thrive in the AI era, creating environments where both humans and AI can perform at their best.

8. It’s not prompt engineering. It’s problem formulation.

The ability to frame the right question will always be more valuable than finding answers. Knowing how to formulate problems worth solving, breaking challenges down logically, and directing attention to meaningful issues are timeless skills that become more crucial in the AI age.

As desk research becomes automated, value shifts to asking better questions. Those who master this meta-level thinking will harness AI’s capabilities most powerfully, turning technological potential into meaningful solutions.

9. Imperfection: from flaw to advantage.

In a future where AI can generate perfect content, human-made creations , with their inherent imperfections and uniqueness , will command a premium. Just as handcrafted pottery became more valued after industrial manufacturing, authentic human creation will transition from utility to artistry, meaning, and connection.
In an ocean of algorithmically perfect content, we’ll choose human-made things not because they’re better, but because they’re human.

10. Future humans: from problem-solvers to meaning-makers.

For millennia, we defined ourselves by our ability to solve problems. AI forces us to find a new identity. As machines take over production and problem-solving, humans will shift toward becoming curators of experience, meaning-makers, and connection facilitators.
In a world where AI handles logic better than we do, we many not need more problem solvers. We’ll turn into meaning makers and beauty recognizers.