In this episode of Nerd @ Work Lab, I sit down with someone I know well: Gaetano Castaldo — strategic technology consultant, former CTO, and long-time colleague. Our conversation dives into one of the most discussed and misunderstood topics of the moment: how small and medium-sized enterprises are approaching artificial intelligence.

The discussion is open, sometimes amusing, and often revealing. Here’s what I took away from it.

Everyone wants AI — but very few know why

I keep meeting companies that “want AI” because their board says so, or because they fear competitors are already using it. Gaetano sees the same trend every day: the pressure to adopt AI often comes from anxiety rather than a strategic objective.

That’s why many pilots end up abandoned. Not because the technology fails, but because no one defined a clear problem to solve. I’ve seen chatbots built simply because they were fashionable, and prototypes launched without anyone assigned to maintain or govern them afterwards.

When the motivation is hype, the outcome is almost always disappointment.

Data comes first — and it’s where most companies struggle

If there’s one theme Gaetano and I are perfectly aligned on, it’s this: AI is a data initiative before anything else.

Yet many SMEs are still catching up on basic digital transformation. I talk to companies where CRM is confused with a “module of the ERP”, or where customer data lives in spreadsheets scattered across teams.

Sometimes businesses ask for predictive systems without having any sensors installed, or forecasting models without historical data. When that happens, the only honest answer is: “First, let’s make sure we have something to predict.”

AI can’t build value if the foundation doesn’t exist.

Where AI truly works

One of the parts I enjoyed most in this conversation is how concrete Gaetano is. We’ve both observed that the most successful AI use cases share clear characteristics:

• They involve large volumes of text or structured data
• They automate or support repetitive human decision-making
• They enhance skilled professionals rather than attempt to replace them

We looked at real examples—from legal automation tools to patent-search engines—that work precisely because they amplify an existing, structured process. When the process is solid, AI becomes a multiplier.

But if the process is unclear, AI simply accelerates the confusion.

Why AI projects fail — and why companies don’t see it coming

I shared a pair of statistics that have stayed with me:

• Nearly half of AI projects fail to reach their goals
• Of those that go live, about 70% don’t generate measurable economic value

This doesn’t mean AI doesn’t work. It means companies rarely know how to measure their processes.

Gaetano actually built a free ROI calculator for leaders who want to self-assess their readiness. I love the idea: you can’t estimate the impact of AI on customer support if you don’t even know the current cost per ticket.

Without metrics, every AI project becomes a leap of faith — and faith is a terrible KPI.

From junior developers to data specialists

A more delicate part of our conversation was the future of junior roles in tech. With AI tools generating code, prototypes, and full application scaffolding, many entry-level tasks are disappearing.

Gaetano believes these roles won’t vanish but evolve. The next wave of junior professionals will be data specialists, people able to understand, clean, organize, and connect the data that modern AI systems need.

Universities, though, are moving slowly. Students often use AI tools daily, while many professors struggle to adapt. It’s a gap that will affect the next generation unless addressed seriously.

A technological turning point — and a cultural one

I said it clearly during the episode: I’m optimistic. Yes, there are ethical issues, organisational challenges, and social risks. But we’re also living through one of the most exciting periods in technology.

It feels like the early days of the web mixed with something much bigger. My own view — perhaps influenced by years of sci-fi — is closer to Star Trek than Terminator. I see AI as a tool that can uplift humanity, provided we use it wisely and critically.

For that, we need data culture, strategic thinking, and the humility to accept that AI is an amplifier — not a miracle.

What comes next

Before closing the episode, we touched on a topic I’ll explore more in the future: the role of schools in preparing students for an AI-driven world. Gaetano is deeply involved in national training initiatives and sees the gap widening between what students can do and what institutions teach.

It’s an essential conversation, and we’ll return to it soon.