A Portuguese artificial intelligence company is taking a different approach to the growing energy crisis created by AI: building sustainability directly into its platform from the design phase rather than trying to fix the problem later. Automaise, which specializes in AI solutions and automation for customer service, is developing the "Automaise Sustainable AI Platform" with backing from Portugal 2030, a government initiative supporting innovation projects. The initiative, which began in April 2024 and is scheduled for completion by September 30, 2026, represents a shift in how companies think about AI development in an era when energy consumption and carbon footprints have become critical concerns.

Why Is Energy Efficiency Becoming a Core Part of AI Design?

The increasing use of large language models (LLMs), which are AI systems trained on vast amounts of text data, has raised serious questions about environmental impact. Rather than treating energy efficiency as an afterthought, Automaise is integrating sustainability and energy efficiency criteria from the earliest stages of development. This approach allows environmental metrics to be woven directly into the model development cycle, promoting more efficient and responsible technologies aligned with the energy challenges associated with growing AI adoption.

The company's strategy reflects a broader recognition that the future of artificial intelligence depends on finding balance between innovation, performance, and environmental responsibility. By positioning sustainability as a core pillar rather than an optional feature, Automaise is attempting to deliver solutions that meet business needs without neglecting environmental challenges.

What Are the Three Core Modules of the Project?

Automaise's sustainable AI platform consists of three technological modules designed to address different aspects of AI's environmental footprint:

• Green LLM Engine: This module reduces carbon emissions associated with pre-training, fine-tuning, and inference of language models through optimization of architecture and reduction of the number of parameters required. Essentially, it makes the AI models themselves leaner and more efficient.
• Data-Centric Sampler: This tool develops more efficient data samples, reducing the amount of information needed to train AI models without compromising their quality and performance. This means training models faster and with fewer computational resources.
• AI Carbon Emissions Calculator: This module quantifies the carbon footprint associated with training and using AI solutions, enabling organizations to monitor and reduce their environmental impact over time.

The vertical nature of Automaise's platform, meaning it operates across multiple layers of the AI stack, allows the company to leverage already processed data to develop new methodologies capable of increasing efficiency and reducing computing resource consumption.

"Since 2017 we have developed our platform with the aim of supporting organizations through increasingly efficient AI solutions. With this project, we continue along that path by integrating sustainability as one of the pillars of our technology. The future of artificial intelligence lies in finding a balance between innovation, performance and environmental responsibility, delivering solutions that meet the needs of companies without neglecting environmental challenges," said Ernesto Pedrosa of Automaise.

Ernesto Pedrosa, Automaise

How Is Automaise Executing This Sustainability Initiative?

The project involves a dedicated team of 12 professionals, including four new hires, and follows an agile methodology structured in five phases of research and development spanning 24 months. This structured approach allows the company to systematically develop and test each component of the sustainable AI platform while maintaining flexibility to adapt based on findings.

• Team Structure: A core team of 12 professionals, with four new positions created specifically for this initiative, ensures dedicated focus on sustainability integration.
• Development Timeline: The project runs from April 2024 through September 2026, providing a two-year window for research, development, and refinement of all three modules.
• Methodology: An agile approach divided into five research and development phases allows for iterative testing and improvement throughout the project lifecycle.
• Market Expansion: The initiative positions Automaise to expand its presence in both national and international markets while consolidating its position as a provider of AI-based solutions.

By completing this project, Automaise aims to consolidate its position as a provider of AI-based solutions while expanding its presence in national and international markets. The company's mission to democratize access to AI across different sectors of activity remains central to its strategy, but now with a stronger emphasis on environmental responsibility.

This development comes at a critical moment when organizations worldwide are grappling with the environmental costs of AI deployment. By building sustainability into the platform architecture rather than bolting it on afterward, Automaise is demonstrating that energy efficiency and AI performance need not be competing priorities. The three-module approach addresses the problem from multiple angles: making models themselves more efficient, reducing training data requirements, and providing transparency into environmental costs so organizations can make informed decisions about their AI usage.

" }