How is civil construction becoming artificial intelligence?

By Patricia Vilas Boas

São Paulo (Reuters) – The use of artificial intelligence agents, especially in the form of virtual assistants, has gained strength with companies focused on residential enterprises promoting efforts to make technology one of the pillars for efficiency gains.

AI agents are autonomous systems capable of performing specific tasks without human intervention. In construction, the adoption of this technology in Brazil has been more focused on areas such as sales, legal and human resources, rather than directly in the construction sites.

MRV, for example, one of the largest residential project builders in the country, began to adopt the Generative AI to train its customers, Mia, which states that about 70% of the group’s sales pass.

Also read:

“In 2023 we became the key, we started looking at the Generative AI just beyond this conversational part,” said MRV’s director of technology and digital transformation, Reinaldo Sima.

The number of those responsible for receiving “leads” – potential customers – at MRV was reduced from 100 to 20 with the support of MIA, the executive said referring to the team that redirects possible customers who pass by Chatbot to the company’s more than 5,000 brokers in Brazil.

With the reduction of this team allowed by the tool, MRV transferred “these people to other activities that are of the highest added value,” he said.

In addition to Mia, the construction company has other AI agents, such as the virtual assistant Susi, the support of brokers Marco and the “best friend of the work”, which supports the construction process. All of these tools are integrated into a platform created in conjunction with Adalink, a Brazilian AI startup created in 2024.

The directional direction began to use AI more broadly by 2024, with applications in customer service through digital and telephone channels, which resulted in an average reduction of 25% in the qualification time of a lead.

Also read:

The group has also used technology in recruitment and selection processes and for support in the legal department.

According to Directional Digital Center Superintendent Eduardo Mocarzel, process automation in different areas of the company has returned more than 10,000 hours of work to the teams monthly.

Since last year the group has adopted AI as “one of the central pillars of its digital transformation journey”, consolidating solutions focused on operational efficiency, customer experience and scalability, according to the executive.

Construtora Tenda, which has been the virtual assistant since 2020, began working on great Language Models (LLMS) with greater emphasis this year, with the launch of a platform that integrates technologies such as chatgpt, Claude and Gemini in a private tent environment, also with Adalink.

“It’s been about a year since we’ve been looking at projects in this regard and it’s been six months since we intensified the agenda in corporate operations,” said Igor Gomes, director of information technology at the Tent.

Tent now works to train its teams to create digital agents within the platform, aiming at developing automated tools that solve everyday problems, Gomes said. The executive highlighted as an example the use of agents for the interpretation of legal documents and the compilation of large volumes of information in the area of ​​Human Resources.

According to the executive, the construction company is also developing an agent for the area of ​​lead up and sales generation.

Also read:

High -end

In the segment of High and Middle Standards, Construtora and developer Eztec launched in October last year the TEC.IA Conversational Tool, to support her real estate brokers Tec Vendas, with about 52% of group ventures registered in the platform database.

Implementation has helped Eztec halve the conversion time in a sales cycle from about a month to 15 days, according to the incorporation superintendent, Tellio Totaro. The next step, according to him, is the development of an application for customer service.

“When (lead) arrives in the broker’s hand, it will arrive with this pre-service done and with more quality … It is a process that should start later this year,” said Totaro.

The LLMs used by Tec.ia are also from OpenAi (ChatgPT), Google (Gemini) and Anthropic (Claude), but unlike tent and MRV, which use a third infrastructure, the development of the platform owns EZTEC.

AI challenges and future

Although the age of AI agents in construction is a reality, it is still in its early stages and its application is limited, especially in smaller builders, according to CBIC’s vice president of Materials, Technology and Productivity Commission, Dionyzio Klavdians.

“Big companies are the ones that are most able to invest in research, technology and advances such as artificial intelligence. The bulk of the construction sector is not able to apply these advances on their own,” he said.

Use in engineering processes and construction sites is also more restricted than in administrative and managerial processes. “These stages of the tip are already developing faster than ‘core’, which is the construction itself,” he said.

“When you go to the site itself, this is another story. Because construction is not yet modularized, not yet assembled, it is still built. And then gets in the way, makes it difficult.”

This perspective is shared by Gomes from the tent. However, the executive states that AI must change this scenario in the future. “If you think of the long run, you can think of a model where you were going to work on automation inside the site.”

The executive cites the expectation of “robotization” of construction, where machines could assist in the production and transportation of structures, but that should only be achieved after a long “digitization journey” in the sector, which “in itself is already a challenge”.

“If you have all this well -structured chain, you can think about the future (EM) a level robotization, perhaps, as a very automated Amazon distribution center. Today, the level of automation in the flowerbed is very low.”