From uploaded project documents, Civils.ai generates answers tailored to construction projects
Civils.ai, the civil engineering focused AI tool that applies a large language model (LLM) to help reduce the time it takes to search for information in construction project documents, has delivered several new updates in its 2.0 release. This includes usability and management improvements and specific new functionality for geotechnical engineering.
The software works by users uploading PDF project documents, such as site reports, contracts or design codes. The AI then processes the data so it can generate answers to specific questions posed by users relating to their project. This could be querying deliverable due dates, through to extracting information on geological conditions from site investigation reports.
The tool is developed by a team led by former Arup and Morgan Sindall civil engineer Stevan Lukic.
According to Lukic, the key difference between his system and ChatGPT (the most well-known example of LLM), is that Civils.ai allows some fine-tuning of responses, whereby site-specific data not contained within the datasets of OpenAI can be temporarily stored in the AI’s memory, using technology known as a vector database.
The user is then able to ask questions and search through the database and retrieve an answer to the question.
“To avoid incorrect or false data being retrieved and misunderstood by the user, the exact page and section of the document is also retrieved and displayed for the user to check and verify against the original document that the answer to the question is correct,” explains Lukic.
According to Lukic, Civils.ai has more than Aecom and Stantec. Major contractors in the UK are also considering the platform.
