You’re Holding It Wrong: Why We’re Missing AI’s Value

When Apple’s iPhone 4 launched, customers complained about poor reception. Apple’s infamous response: “You’re holding it wrong.” The problem wasn’t the phone, they said — it was the user.

Today, I find myself feeling the same way about AI. The technology is powerful. The opportunity is real. But most organizations are holding it wrong.

The Golf Program Example: Capability Meets Context

A friend of mine recently used Cursor AI to build a program for managing golf tournaments. He’s not a professional developer — he’s an engineer with a passion for golf. However, by combining his systems engineering mindset with a conversational coding assistant, he built a working application in just a few weeks.

AI didn’t make him a developer. It 10Xed his ability to execute because he already knew how to think in systems — how to break down a problem into logic, data, and workflow.

This is the key pattern: AI amplifies technical intuition. It collapses the communication gap between idea and implementation when the user can already think like an implementer.

IBM’s “Bob” Platform: When AI 10Xs Developers

IBM’s Project Bob is a prime example of AI held right.

As IBM’s Dinesh Nirmal shared at TechXchange, 6,000 developers across IBM are now using Bob and seeing 45% productivity gains — not because Bob replaces them, but because it augments their ability to complete complex modernization and remediation tasks.

Neel Sundareshan, IBM’s GM of Automation, explained that Bob’s success wasn’t about retraining users. There was no manual. Developers simply started using it — and 95% of their work shifted to AI-assisted tasks.

These figures from IBM’s TechXchange underscore a principle I’ve emphasized repeatedly: AI’s value is unlocked when it’s embedded directly into the workflow of the technical expert. When the model is close to the work, value emerges naturally.

The 5% Problem: Capability Without Usability

Nirmal opened his keynote with a reality check: only 5% of enterprises are seeing measurable value from generative AI. He explained why:

“Infusion of generative AI has two pieces — availability and usability. Just because something is available doesn’t mean it’s usable.”

That’s the industry’s “You’re Holding It Wrong” moment.

We’ve made AI available — thousands of models on Hugging Face — but we haven’t made it usable for the people closest to the problem.

The Cloud Migration Example: Capability Without Compression

Look at what’s happening in cloud migration today.

Moving from VMware vSphere to another platform — or from a private data center to a hyperscaler — isn’t a technical impossibility. It’s a human bottleneck.

Every migration involves thousands of hours spent translating business logic into system architecture: dependency mapping, configuration, optimization, and validation. The technical problem is solvable; the translation cost is what kills velocity.

IBM’s keynote highlighted that 59–68% of enterprise software development is modernization, not innovation. Most of our engineering capacity is spent rewriting what already works — not creating what’s next.

Now imagine AI-driven agents that could take an existing vSphere environment and automatically refactor it into a “scale-to-zero” architecture in the cloud — maintaining function while transforming form.
(Scale-to-zero means the resource consumes no cost when not actively running, a key efficiency driver in the cloud model.)

That’s not a developer challenge anymore. It’s a pattern recognition and decision constraint problem — exactly the kind of work AI is built to handle.

The challenge is putting that capability in the hands of the teams that actually understand the systems — the people who know what needs to move and why.

Why Non-Technical Users Don’t See the 10X

Here’s the hard truth: AI 10Xes people who already think like engineers. Developers, architects, and system admins see the lift because they can already express their intent in structured ways.

But when non-technical users — business analysts, project managers, or operations leaders — try the same tools, the amplification collapses.

As Anthropic CEO Dario Amodei said in his IBM discussion, enterprises “are filled with experts in their domains, not experts in AI,” and their existing processes and governance create friction for adoption.

We haven’t failed to train people; we’ve failed to design systems that speak their language.

The Real Lesson: Redesign the Antenna

Apple didn’t retrain millions of iPhone users to hold the device differently. They redesigned the antenna.

The problem wasn’t the user — it was the design.

That’s exactly where we are with AI.

We don’t need to teach every business unit how to “prompt better.” We need to redesign AI interfaces so they fit the way business and IT professionals actually work.

That means embedding AI inside workflows — not on top of them. It means giving infrastructure teams AI agents that can act safely in production, and process owners tools that automate modernization, not just describe it.

Final Thought

AI’s promise lies in collapsing silos — between business and IT, between intent and execution.
The winners won’t be the organizations that deploy the most models, but those that redesign the experience of computing itself — embedding AI agents directly into the workflows of the engineers and architects already positioned to execute.

That’s how you hold it right.

Speaking of holding it right, have you tried my AI project? It’s my AI persona trained on some 7K pieces of content – Virtual CTO Advisor.

Ask it how to overcome some of the barriers to successful AI deployment.