Darwoft at Data Science & AI Fundamentals IES Conference Series · Opening Talk

“AI empowers you, it gives you superpowers, but only if you can control it.” — Sheila Cravero, CAIO, Darwoft

On Tuesday, September 29, Sheila Cravero, Chief AI and Innovation Officer, and Santiago N. Porta, Co-founder and CTO of Darwoft, opened the IES Conference Series with the talk Data Science and AI Fundamentals. Speaking to students at Colegio Universitario IES in Córdoba, Argentina, and to an audience following the YouTube livestream, they showed how a software company works with AI today: with clear rules, constant measurement, and people in charge of every decision.

Key topics of the talk

The session combined strategy, technical concepts, and a live demo. These were the main topics:

• Why governing AI is safer than banning it.

• What the ISO 42001 standard requires for responsible AI use.

• How AI agents are built with context, skills, and MCP.

• Which skills will set tech professionals apart.

Governing AI: the brakes that let you accelerate

Sheila explained that many organizations believe not using AI avoids risk. In practice, the opposite happens: people use it anyway, and without rules, the chance of data leaks grows. That’s why Darwoft chose to govern it, focusing on three fronts: protecting information through licenses and confidentiality agreements, training everyone in safe use, and measuring how AI is used.

That measurement, done through telemetry, helps detect risks, manage token costs, and evaluate results with two metrics side by side: speed and quality.

To explain the ISO 42001 certification, Sheila used a simple analogy: if AI were a car, the standard would be the driver’s license. She highlighted three of its requirements: defining and testing tools (for security, reliability, and bias), establishing methodologies with controls, and recognizing that final accountability always rests with a person. That’s why all of Darwoft’s work includes a human in the loop.

From chat to AI agents

Santiago took the talk into technical territory and showed what lies behind the chat window everyone knows. He walked through the concepts Darwoft uses today to build software with AI:

Context, an index file that tells the AI the project’s architecture, rules, and documentation. Skills, files with instructions that guide how each task should be solved. MCP, a protocol that connects AI to other tools, such as Figma, so it can retrieve information and take actions.

These pieces are used to build agents. Santiago explained it with an everyday example: a mom who hands out household chores, checks how they turned out, and asks for anything unfinished to be redone. That’s how an orchestrator works with its agents and a reviewer. Breaking a problem into small parts, each with a focused context, improves the quality of the result.

He also showed how to set boundaries: permissions that define what AI can and can’t do, automatic checks before actions run, and isolated sandbox environments.

A live demo

To wrap up, Santiago presented an example of four agents working together: a scriptwriter who creates a superhero, a villain agent who finds an enemy, a designer who creates the suit, and a reviewer who scores the result. The entire application was built with AI. Santiago didn’t write a single line of code, but he defined what he wanted to achieve, how it should work, and what criteria to use when evaluating it.

Takeaways from the Q&A

Questions came from both the room and the YouTube chat. Some ideas that stood out:

The key choice isn’t the model, it’s the working methodology. A good practice is having the reviewer use a model from a different provider. Each agent can run on a different model depending on the task and the cost. And Data Science keeps growing, for example in the US healthcare sector, where data analysis is central.

On professional growth, the advice was clear: build strong general foundations, specialize in something, and add an understanding of business, architecture, costs, and soft skills. Creativity remains on the human side.

If you’re taking your first steps with AI or looking to go beyond the chat window, this talk is a great place to start.