Leonardo Felipe Nerone Says Enterprise AI Has an Implementation Problem

Inside many companies, AI is already present, but the work still feels strangely unchanged. Teams have access to new tools, internal documents, dashboards, databases, spreadsheets, and software systems. Yet employees still spend hours searching for answers, repeating manual tasks, and moving information from one place to another.

For Leonardo Felipe Nerone, co-founder and CTO of Drachma, that disconnect says a lot about where enterprise AI stands right now.

“Most companies do not have an AI problem,” Nerone says. “They have an implementation problem.”

That is the problem Drachma was built to solve. The New York-based company builds custom AI systems, assistants, and agents for businesses. Drachma is not built around one-size-fits-all AI. It connects artificial intelligence to a company’s actual data, APIs, internal tools, permissions, workflows, and business rules. The goal is not to give employees another place to type questions. It is to make AI useful inside the way the company already operates.

Nerone believes the market is entering a more serious phase. For the past few years, many businesses have treated AI like something to test, announce, or place beside existing systems. That phase produced interest, but not always impact. A prototype may look polished in a controlled presentation, but fail once it meets the company’s real environment.

“A controlled presentation can hide almost everything that matters,” he says. “It does not show whether the system understands permissions, whether it can reach the right data, whether it fits the workflow, or whether people will actually use it.”

That is where Nerone sees the next opportunity. Companies are no longer asking only whether AI can answer questions. They are asking whether it can complete parts of workflows, retrieve information safely, analyze documents, use tools, and support decisions across teams. That requires more than a strong model. It requires system design.

“The model is only one part of the system,” Nerone says. “The real work is context, integration, security, reliability, and business outcomes.”Drachma’s work often begins with a practical question: where is the business losing time, speed, quality, or money?

In some companies, the issue is scattered knowledge. In others, it is repetitive internal work. Some teams need AI features inside their own products. Others need an internal assistant that can help employees navigate complex data without depending on manual analysis or technical support.

Nerone argues that the wrong starting point is asking which AI tool a company should buy. A better question is which business process should be redesigned.“A generic AI tool can help an individual be more productive,” he says.

“But company-level impact usually needs something deeper. You have to understand the process, the data, the permissions, and the people who will use the system.”

Nerone’s view is shaped by years spent building across payments, fintech, crypto, open-source systems, data environments, and enterprise software before co-founding Drachma. That range matters because he is not interested in artificial intelligence as a surface-level feature. He is interested in the point where AI leaves the test environment and has to fit the company’s actual operating rhythm.

“Good AI systems are not built only around what the model can say,” he says. “They are built around what the business needs to happen.”

That distinction has become especially important as companies talk more about AI agents. In theory, agents can plan, use tools, access data, and complete multi-step workflows. In practice, Nerone believes the industry often uses the word too loosely.

“Not everything called an agent is really an agent,” he says. “Sometimes it is a chatbot. Sometimes it is automation. Sometimes it is a simple assistant with a new name.”

The deeper issue is not terminology. It is whether the system is appropriate for the workflow. Nerone does not believe every business process needs full autonomy. In many cases, the best AI system is not the one that does the most by itself. It is the one that does the right amount safely.

“More autonomy is not always better,” he says. “For some workflows, AI should prepare information. For others, it should recommend an action. In some cases, it can execute. The important thing is matching the level of autonomy to the risk and the value of the task.”

That is why trust becomes part of the product. Once AI connects to internal systems, companies care about permissions, logs, data boundaries, human review, security, and reliability. A system that ignores those concerns may feel faster at first, but it will not last inside a serious business environment.

Drachma’s approach is to build around the client’s operation from the beginning. That can mean connecting to databases, reading internal documents, working with APIs, respecting permission structures, and designing interfaces that fit how employees already work. The company has built tools such as a custom chatbot connected to a client’s data environment, allowing users to interact with complex information in natural language instead of relying on manual analysis.

The value, Nerone says, is not simply that AI is present. The value is that work changes.

“AI should reduce friction,” he says. “It should help people find information faster, make fewer mistakes, and spend less time on repetitive work.”

That view also shapes his advice to technical founders and companies entering the space. Start small. Choose one workflow. Define a baseline. Test the system in the real environment. Measure the result. Expand only after the value is clear.

“An impressive prototype is not the same thing as a useful product,” Nerone says. “The difference is usually in the details that users never see: data engineering, integrations, security, evaluation, context, and reliability.”

After pivoting from an earlier payments idea, Drachma has spent about a year focused on custom AI systems for businesses. Nerone wants the company to become a lean, technically excellent business known for making AI practical inside companies in the United States and Brazil.His vision is not built around replacing people. He sees AI as a way to reduce wasted effort and help teams operate with more leverage. For information on Leonardo Felipe Nerone, visit his LinkedIn.

“In many cases, the greater impact is helping people become more capable,” he says. “Faster analysis, fewer errors, better decisions, and more time for higher-value work.”

For Nerone, that is where enterprise AI becomes real. Not in the announcement. Not in the controlled presentation. Not in the label attached to a tool. It becomes real when it changes how work gets done.“AI will matter when it is connected to the reality of the business,” he says.

“That is the hard part, and that is the part worth building.”