Now the Work Begins
Romero AI and Souto AI point to a new phase of artificial intelligence: less technology hunting for a use case, more real problems finding a way to get solved.
For a while, artificial intelligence had a science fair phase.
Every week brought something that could write, draw, code, talk, sing, summarize a meeting, or produce in 30 seconds what used to take an afternoon.
It was hard to look away.
But every consequential technology eventually reaches a curious milestone: it starts becoming less interesting.
That is excellent news.
It means we are moving beyond “What can this technology do?” toward the question that actually matters to a business:
“What problem can it solve?”
Digital minds may be reaching that point.
Two projects recently launched in Brazil make the shift tangible. One puts nearly 30 years of Romero Rodrigues’s entrepreneurial experience within reach of people who would struggle to get a meeting with him. The other makes more than three decades of Rafael Souto’s expertise in careers and professional development available to people inside companies.
Romero AI and Souto AI matter because they address two bottlenecks that existed long before anyone knew what a prompt was.
The digital replicas are the delivery mechanism. The problems are the story.
The Best Test Wasn’t Whether
the AI Sounded Like Romero
At the launch of Romero AI, Brazilian business publication NeoFeed had a particularly good idea: ask Romero Rodrigues and his digital mind the same five questions.
The obvious temptation was to stage an impersonation test.
Romero said this. The machine said that. Which answer sounds more like Romero?
But the comparison revealed something more useful.
Asked about his most important decision while leading Buscapé, Romero talked about people, team, and culture. His digital mind reached for another part of the story: the sale to Naspers and his subsequent departure from the company.
Some answers converged. Others shifted emphasis, retrieved different episodes, or followed different lines of reasoning.
Then Romero summed up the experience:
“Its memory is much better than mine.”
That observation gets close to the point.
The value lies in putting nearly three decades of accumulated experience to work on a specific question.
Building Romero AI involved recovering emails from the Buscapé years, interviews, documents, reference material, and even the company’s first business plan.
Romero does not have to keep that entire archive in working memory every time someone asks for advice.
The machine can retrieve.
Romero can judge.
Those are different capabilities. We may have spent too long debating which should replace the other when the opportunity lies in combining them.
Experience Has Terrible Unit Economics
Another detail in Romero’s case matters even more to me.
Entrepreneurs ask him for guidance all the time.
These are not exclusively startup founders surrounded by venture funds, accelerators, and fashionable coffee shops. He mentions bakery owners, people running small chains of gyms, entrepreneurs building real businesses who simply could not get access to him.
The constraint was always distribution.
Experience has terrible unit economics.
It takes 20 or 30 years to build. Once you have it, you still deliver it largely one conversation at a time.
You can write a book. Record a course. Give a keynote to a thousand people.
But the moment someone asks a specific question, “I run this business, in this market, with this problem. What would you do?”, the whole system runs into its scarcest resource.
A slot in someone’s calendar.
That is where a digital mind starts looking like infrastructure.
The Same Bottleneck,
Inside the Company
Now replace the entrepreneur with someone working inside an organization.
They have been promoted and are unsure how to handle the new role.
They want to move into another field.
They are weighing a job offer.
They have received difficult feedback.
They need to work out whether the problem is their job, their manager, their career, or themselves.
There is plenty of content about all of this.
What is harder to distribute is guidance that takes their actual circumstances into account.
That is the problem Produtive has begun addressing with Souto AI, a Cognitive Extension built on more than 30 years of Rafael Souto’s experience and a knowledge base of over 250 pieces of content covering careers, leadership, employability, and professional development.
Its departure from conventional employee development starts with a simple observation:
Career dilemmas do not follow the training calendar.
They show up when they show up.
Once again, we encounter the same bottleneck: valuable knowledge on one side, demand on the other, and a limited supply of people, hours, and money in between.
Romero AI and Souto AI address different problems. Economically, though, there is an interesting symmetry.
One Helps Build Companies.
The Other Helps Them Work.
Entrepreneurs are one of an economy’s engines.
They turn problems into products, jobs, companies, and markets.
Professionals are another.
They turn capital, technology, and strategy into execution inside organizations that already exist.
Romero works at one end of that equation, helping people build businesses.
Souto works at the other, helping people build their careers and, in turn, strengthen the companies they work for.
One helps bring new businesses into existence.
The other helps existing businesses put better decisions into practice through their people.
At MAXMEAI, we had the opportunity to implement both digital minds.
Naturally, I am proud of that work. But the most encouraging part, for someone working in technology, is watching the technology begin to disappear behind the problem it solves.
The most interesting thing about a digital mind is usually the bottleneck it removes.
We Have Enough Names.
Now Show Us the Spreadsheet.
Over the past few years, we have developed a vocabulary for this emerging layer.
Digital clones. Digital humans. Cognitive Extensions. Infinite Knowledge Legacy. Eclonomy.
The concepts helped us recognize what was taking shape.
Eventually, though, concepts have to leave PowerPoint and survive a spreadsheet.
Two years ago, a convincing AI demo was news.
Today, it should be onboarding.
If a digital mind supports entrepreneurs, we should ask whether it helps someone make a better decision, find a faster route, or avoid an expensive mistake.
If it supports professional development, we should ask whether people use it, come back, make better decisions, develop skills, and improve something the company was already trying to improve before AI arrived.
In sales, measure sales.
In customer support, resolution.
In education, learning.
In productivity, time.
Reducing a fascinating technology to these questions may sound brutal.
That is how fascinating technologies become important ones.
AI Is Moving Beyond the Search
for Something to Do
Perhaps this is the real change.
During the early years of generative AI, companies asked the same question in dozens of ways:
“Where can we put AI?”
Understandable.
Give someone an extraordinarily powerful hammer, and they will start looking for nails.
Now the question is beginning to turn around.
Where is there an expensive, recurring, poorly solved problem because a valuable human capability cannot reach enough people?
Where does demand exceed the availability of experts?
Where does a good decision depend on experience concentrated in a handful of minds?
A digital mind may make sense there.
That reversal matters. It puts the problem back at the beginning of the process.
The less futuristic these applications look, the more interesting they become.
Romero AI does not have to transform the future of entrepreneurship. If it helps a small business owner avoid a decision that would cost six months and money they cannot afford to lose, it has created value.
Souto AI does not have to reinvent the future of work. If it helps someone approach a difficult career decision better prepared, it has changed something that matters.
Now multiply that by thousands of interactions.
The conversation moves from technological curiosity into economics.
Perhaps the next AI race will be about who finds the right problem to put all that intelligence to work on.
I would much rather watch that race than another contest over who has the most powerful model.
The novelty has worn off.
Good.
Now the work begins.