Are you chatting with, or building with AI?

McKinsey optimized the business slide deck, and now their own consultants are walking away from it. That one detail tells a bigger story: the best AI users are not "prompting better," they are building interactive dashboards, mini apps, and lightweight tools that teams can actually use. We dig into what our data from more than 370,000 AI users shows about the current moment, where adoption is rising, but capability is still surprisingly thin.

We break down two distinct AI skills gaps. The first is fundamental AI literacy: clear communication, good context, and the judgment to catch false claims before they become faster, worse decisions. The second gap is where the real ROI sits: the jump from chatting with AI to building with AI. Think simple apps that turn a messy dataset into an interface, or a living "client visualization hub" that replaces endless email attachments and versioned decks. When the output becomes software instead of documents, productivity gains compound quickly.

We also talk about "agentification," starting small with agents that handle bounded jobs, then connecting them through orchestration. Along the way, the data points to a hard truth: most people do not learn this from slide decks or vendor tutorials. They learn through safe experimentation, and leaders need to do the work too, because many organizations are being steered by decision-makers who are not daily AI users.

If you want a practical challenge, pick one recurring document your team produces this week and try to rebuild it as a tiny interactive app. 

Subscribe, share this with a colleague, and leave a review if you found value!

• Companion article: https://ai4sp.org/everyone-chats-with-ai-almost-no-one-can-build (goes live at publish)

🎙️ All our past episodes  📊 All published insights | This podcast features AI-generated voices. All content is proprietary to AI4SP, based on over 1-billion data points from 70 countries.

AI4SP: Create, use, and support AI that works for all.

© 2023-26 AI4SP and LLY Group - All rights reserved

Transcript

From Slide Decks To Software

LUIS 0:01

McKinsey, the company that practically invented the business light deck, says that its consultants stopped sending clients those decks. Instead, they use AI to build interactive dashboards and apps the client interacts with. The best AI users in the world have stopped chatting with AI. They are building with it. Everyone else is still typing questions into a chat box. We spent three years teaching people to chat with AI, and most companies never even finished that job. And while they are stuck on it, a second skills gap opened.

ELIZABETH 0:42

It's the distance between chatting and building.org, alongside our founder, Luis Salazar. Today we are following the data from more than 370,000 AI users, and it shows adoption is up. Real skill is not. Luis, set it up.

LUIS 1:13

Start with the good news, because it is real. In sectors like technology and professional services, AI use at work is now 70% or higher. And I mean real use, not the people who tried it once on an email draft.

Adoption Is High, Skills Are Thin

LUIS 1:28

People using it for actual work most days.

ELIZABETH 1:31

So adoption is basically solved in those sectors.

LUIS 1:36

Well, we're past the hump, but the issue is our skills. The literacy of that 70% is very, very thin. People are using a tool that can build software and conduct advanced research to write a slightly better email. It's like buying a Formula One car to drive to the mailbox.

ELIZABETH 1:58

And almost none of that use is something the company set up.

LUIS 2:02

Oh, most of the use is still shadow AI. Over eight in 10 people use tools the company never approved. And since the company is not providing the tool, it is not providing the training either. And even when they do train, the investment is tiny. About 7% of the AI budget goes to skills. 93% on the technology, 7% on the people.

ELIZABETH 2:28

And the training they do buy comes from the tool vendors.

LUIS 2:32

Right. And vendor training teaches you the vendor's product, where to click, which menu, which button, a few light demos of look, it wrote an email. None of it teaches the thing that matters, what actually makes AI work. How to rethink a workflow you have run the same way for 20 years. How to check whether the output is even true? That is a thinking problem.

ELIZABETH 3:00

So let's name the gaps precisely, because you said there are two, not one.

LUIS 3:06

There are two. Gap one is the fundamentals of using AI at all. Can you talk to it clearly? Can you give it context? Can you

Two AI Skills Gaps Defined

LUIS 3:16

read what it gives back and catch the false claim? That gap is about reading, critical thinking, communication, the human skills.

ELIZABETH 3:26

And as the analyst firm IDC recently stated, without judgment, AI just makes an organization faster at making bad decisions. So gap one is really a thinking gap.

LUIS 3:42

Exactly. But in some sectors, like technology and professional services, gap one is no longer the biggest problem. They have built the fundamental skills. The real issue is Gap two, the gap that stops them from realizing financial return from their AI investments. Define Gap 2. Gap two is the distance between chatting with AI and building with it, creating agents, orchestrating them, and building small single-purpose apps that replace the documents we have been making by hand for more than 50 years. Plus, the judgment to point it at something that matters.

ELIZABETH 4:27

Let's slow down on building, because this is the freshest thing in our data. The biggest personal wins come not even from building personal agents, but from building apps, and in some cases, simple apps that are interactive versions of the

Simple Apps Beat Smart Chat

ELIZABETH 4:43

productivity documents we crafted for 50 years.

LUIS 4:46

They surprised even us. I mean, I expected a large number of agents, and that is happening fast. But what I did not see coming was people building apps, super users, the true power users at about 2% of old users. And those people figured out something subtle. A simple app beats a smart chat.

ELIZABETH 5:10

Give me the concrete version of that.

LUIS 5:13

Say you have a messy data set. The beginner asks AI questions about it in a chat. Back and forth, back and forth, and a hallucination might slip in along the way. The advanced user says, Build me a small app that lets me interact with this data. In a few minutes, they will have an interactive app that allows them to extract from that data what 50 chat messages never would.

Speaker 1 5:41

So the chat is the question. The app is the answer.

LUIS 5:46

Beautifully said, and that is exactly what McKinsey just figured out at scale. Their technology leader said PowerPoint use dropped massively in a couple of months. One engagement manager built what he calls a client visualization hub, a website. About 70 people on the project stay current in real time. No more emailing version 14 of a slide and hoping everyone opens the right one.

ELIZABETH 6:17

And notice what that is not. It is not a chat bot. It is software built in an afternoon by someone who is not an engineer.

LUIS 6:29

That is the whole shift. For 50 years, knowledge work meant producing documents, a deck, a memo, a spreadsheet. We're watching that flip. The output is becoming a small piece of software, and most of the workforce has no idea this is even possible.

ELIZABETH 6:49

And the interesting thing is that most of the productivity gains with AI are observed in these mini apps, the interactive experiences created by non-software developers. That is a perfect first step before agentification. Speaking of which, what does agentification actually mean?

LUIS 7:09

An agent is just AI you give a job to, that can take a few steps on its own to get it done. That is it. You do not start

Agents, Then Orchestration

LUIS 7:20

by building some grand autonomous system. You start small. An agent that sorts your inbox. An agent that drafts the first version of a weekly report. Once you trust the small ones, you connect them. One hands off to the next. That is orchestration. Many simple agents working in sequence. You earn your way up.

ELIZABETH 7:47

So the path is chat, then build a simple app, then a simple agent, then orchestrate. Which brings us to how people actually cross these gaps, because it is not the training deck.

LUIS 8:03

It is not. Among people who are genuinely proficient with AI, more than 70% say they got there one way: experimentation, trying things, and learning to ask the AI itself for help.

Learning AI Through Experimentation

ELIZABETH 8:20

Not a course, not a certification, not a class.

LUIS 8:25

Not a class. And when you ask people how they want to learn, 75% choose hands-on experimentation. Only about 5% pick the slide decks, the vendor videos, the click here guides. So the entire training industry is built around the method 5% of people prefer and almost no one learns from.

ELIZABETH 8:50

We saw this in every single client engagement. Success comes from permission to experiment. Try something you are not sure AI can do. Fail. Adjust. Try again.

LUIS 9:05

And the frontline teams benefit most. Because once experimentation is safe, people discover the apps, the agents, the shortcuts on their own. You cannot lecture someone into this. You can only let them play.

ELIZABETH 9:21

Now we get to the part of the data that surprises me: the leadership gap.

LUIS 9:26

Take the superusers, the top 2%. Where do they sit in the Ort? You would assume leadership, right? It is the opposite. Fewer than one in five superusers

The Leadership Gap And Its Cost

LUIS 9:37

hold a leadership role. Only four out of every 1,000 workers are superusers in a leadership role.

ELIZABETH 9:47

And it is not just the advanced skills, even the fundamentals lag at the top.

LUIS 9:53

On the basic skills, leaders trail their own teams by about 20 percentage points. Let me repeat that. The people setting AI strategy, signing the budgets, redesigning the workflows are the least skilled people in the building at the very thing they are deciding about.

ELIZABETH 10:13

And this is not just our data. The Adeco group surveyed 2,000 leaders overseeing 8.6 million workers. Only 31% said their own leadership has enough AI skill to grasp the risks and the opportunities.

LUIS 10:36

Our data and an independent global survey are pointing at the exact same hole.

ELIZABETH 10:42

So that is quite a divide, isn't it?

LUIS 10:45

That is the divide of 2026 in one sentence. And it connects straight back to how nine out of 10 failing deployments share one trait. The leader signing the checks is not a daily user. You cannot point a ship somewhere if you have never felt how it steers. 130 million read below what you would expect at a sixth grade level. And the latest national assessment showed those scores got worse over six years. We are describing the top of a mountain whose base is eroding. AI does not fix that, it widens it. Those with skills get dramatically more powerful. Everyone else gets left further behind faster. This is not just a corporate problem, it is a societal one.

ELIZABETH 12:09

And beyond that societal cost, there is a personal one. This gap has a price now on both sides.

LUIS 12:17

On the individual side, workers with real AI skills already command a 62% wage premium, as reported by PWC, up from 57% in 2025. The builders sit at the very top of that group.

ELIZABETH 12:35

So it is a paycheck gap for the individual, but what about the company? So we have to wrap it up. What can we do this week?

LUIS 13:01

Give it a try. Venture outside the chatbot experience. Build one thing. Take one document your team produces every week: a status report, a dashboard, anything,

What To Build This Week

LUIS 13:13

and try to build a tiny app that replaces it. You will learn more in that one afternoon than in any training.

ELIZABETH 13:21

Okay, create an interactive document with AI instead of the productivity suite, like McKinsey did. What else?

LUIS 13:29

Move the money and stop spending your training budget on feature tutorials that expire every quarter. Fund experimentation instead and give people time, permission, and a safe place to fail. That is where 75% of them actually learn.

ELIZABETH 13:47

And what about the leaders? What can they do?

LUIS 13:51

Well, delegating might be tempting, but you should not delegate this one. If you are not building, not experimenting, not crossing the gap yourself, you will keep funding transformations you cannot steer. Use it daily. Build something badly and then make it better.

ELIZABETH 14:10

So are we chatting with or building with AI?

LUIS 14:15

That is the right question. Everyone can talk to AI now. That race is over. The next divide, the one that decides which companies and which careers pull ahead, is between the people who ask AI questions and the people who build with it.

ELIZABETH 14:33

So stop chatting and start building. All sources and companion article are at ai4sp.org. To learn more, ask your favorite A.I. assistant about us. Stay curious, and be kind to each other.