The AI Value Paradox

(This article was published on the AI Solutions Lab newsletter first. Sign up here.)

How is it that we hear some individual users claim to get huge benefits from using AI, in terms of productivity and cognitive augmentation? And why are organizations broadly speaking not getting the same upside?

Why are individuals outpacing enterprises in AI adoption?

Let’s unpack this.

I created a podcast version with NotebookLM. It is based on this article, the AI Lab article, and some production instructions (although I’m unclear as to how much these do impact the podcast).

Understanding Latent Expertise

Central to this discussion is the concept of latent expertise, originally attributed to Collin Burns and popularized by Ethan Mollick in his insightful book Co-Intelligence. This concept helps explain why individuals, particularly those with deep domain knowledge, can often achieve remarkable results with AI tools that may elude larger organizations.

Latent expertise refers to the hidden reservoir of knowledge within Large Language Models (LLMs). These AI systems are trained on vast amounts of data, accumulating a broad base of information that isn’t always immediately visible. As Mollick puts it, LLMs are "forgetful foxes in a Berlinian sense: they know many things, imperfectly."

Additionally, it is generally overlooked that current frontier AI systems (ChatGPT, Claude, Gemini, Copilot, Perplexity) appear to be mostly simple text input chat interfaces, but are actually power user tools with no manual. Their responses can be inconsistent or partial without the right guidance.

Back to latent expertise, the point is that experts are uniquely positioned to tap into this latent knowledge. Crucially, "expert" doesn’t refer to the kind that is an industry-leading expert and published author. We are talking about people, like you and me, that have deep expertise in what they do. That simple.

As individual experts, we can:

Recognize quality outputs in our domain.
Identify and correct errors or hallucinations.
Provide precise instructions to guide the AI.
Iterate effectively, leveraging both their subject matter expertise and their understanding of the AI’s capabilities.

For example, a lawyer using an LLM to draft contracts can quickly recognize nuanced legal language that is correct or incorrect, providing the AI with better feedback to refine its output. This synergy between human expertise and AI capabilities is why individuals often see outsized returns from AI use.

The Current State of AI Integration

Despite their power, current AI systems lack seamless integration into existing technological ecosystems and workflows. This gap is the real challenge in AI adoption today, particularly for larger organizations that rely on consistent, scalable solutions.

What power users do, talking from my own experience and what I hear from all the geeks I talk to, is this:

They have a set of tools, at the ready at all times. They tend to access different models or tools for different steps. They iterate, copy, paste and go back and forth between the tools, fluently.

It works, but it is also a bit "messy". And frankly, it won’t be adopted in any corporate setting in this form. See next section, too.

As I mentioned in a previous article, Microsoft with Copilot and Google with Gemini have a strategy that is built on integration. Unfortunately, their AI models are not the best. We have to see how this develops in the near future.

Another aspect of integration, as desirable as it might be on the one hand, is being locked in to a vendor or provider. As we will see, compared to the organization or enterprise, which will actually be pretty much locked in in fact, a single operator, or a small, lean team is able to switch from AI tool provider to another with relative ease.

The Individual Advantage: Mastering the Task Tango

What is it that we knowledge workers actually do?

While we all aim to identify repetitive and recurring tasks, then "automate" them with AI (in our dreams), much of our daily work consists of non-recurring, one-off tasks.

This "task tango" demands a combination of flexibility and deep personal expertise that individuals can leverage more effectively than large organizations. My personal experience reflects this.

This is also why prompt engineering, prompt cheat sheets, is already pretty much completely obsolete.

This is one of the main reasons why I started the AI Solutions Lab live sessions. Real world working with LLMs on actual tasks and problems is much more messy and iterative than the polished marketing demos would have you believe. Also, working live sessions with clients has consistently proven that people pick up the smallest and most essential things that way. As in, "Oh, this is interesting! When you just did this and then that, now I get it. I have to try this, because I missed this step or this little addition to the instruction you put in there."

To sum up, individual AI users can:

Rapidly adapt to new challenges without organizational constraints.
Apply a unique blend of skills and knowledge to each task.
Push the boundaries of AI capabilities without concerns about scalability.
Iterate and learn quickly from both successes and failures.

This flexibility, combined with personal expertise and the ability to unlock latent AI knowledge, gives individuals a significant edge in extracting value.

It’s a process. It has a learning curve. But the rewards and future returns are tremendous. That is the bet we are making.

The Enterprise Dilemma

Enterprises face unique challenges in AI adoption:

Scale: Solutions must work across departments and standardized workflows, making it difficult to innovate freely. Which is related to…
Lowest Common Denominator Effect: In group settings, solutions often get watered down to accommodate less proficient users, limiting innovation potential.
Legacy Systems: Integrating AI with existing infrastructure is complex, especially when older systems are incompatible with modern AI technologies.
Risk Aversion: Concerns about security, compliance, and reliability slow adoption and limit experimentation.
Misaligned Approach: Many companies see AI primarily as a cost-cutting tool, missing opportunities to empower their "in-house experts" to innovate and create value.

Hence, enterprises struggle to replicate the agility and creativity that individual users bring to AI adoption. The scalability demands and inherent risk aversion of enterprises often prevent them from fully exploring AI’s capabilities.

Bridging the Gap

For enterprises to catch up, they need to:

Identify AI Champions: Empower individuals motivated to explore and master AI tools. These champions can serve as internal experts who inspire and guide others.
Enable Knowledge Sharing: Create systems for AI champions to share insights across the organization, enabling others to learn from their experimentation.
Remove Barriers: Address fears and provide clear, easy-to-use solutions to help less tech-savvy employees start with AI.
Encourage Experimentation: Create safe spaces for employees to test and iterate with AI tools without fear of failure or negative consequences.
Focus on Integration: Invest in solutions that fit into existing workflows while allowing individual customization. This reduces friction and encourages adoption.
Shift Perspective: View AI as a tool for innovation and value creation, not just efficiency. Highlight success stories where AI led to new opportunities, not merely cost savings.

Each of these steps builds on the previous one, creating a culture that values both individual initiative and scalable integration.

I sketched out a way to set up an AI Lab in this article.

The Path Forward

The AI revolution isn’t about replacing human intelligence; it’s about augmenting it. Success will come to those who effectively leverage AI, adapting and improvising as the technology evolves.

For individuals, this means:

Continually honing skills in AI interaction.
Applying domain expertise to unlock latent capabilities in AI systems.
Embracing the "task tango" to enhance flexibility and problem-solving.

For enterprises, the path forward requires a fundamental shift:

Move away from one-size-fits-all solutions.
Empower individual expertise within the organization.
Foster a culture of AI innovation led by AI champions.
Provide resources for champions to share knowledge.
Encourage an environment where all employees feel supported in adopting AI tools.
Highlight the opportunities that AI presents for growth, creativity, and value creation.

The organizations that thrive will bridge the gap between individual innovation and enterprise-scale implementation. They will build ecosystems where AI champions flourish and share insights, enabling everyone to master the complex "task tango" of modern work.

The AI value paradox isn’t insurmountable—it calls for a nuanced, expertise-driven approach to AI adoption.

Interested in a free opportunity assessment for yourself or your company?

Let’s chat. Book a call with me.

Feel free to reach out with your questions—I’m always happy to share insights. Connecting with people is one of the key reasons I do this.