UpCodes launched an AI-powered building code research assistant known as Copilot last week, according to a press release.
The launch comes on the heels of a $3.5 million series A funding round led by Building Ventures. CapitalX and Bragiel Brothers also participated in the round, and the funding will fuel hiring for its engineering department and help expand its code library and resources.
Founded in 2016, UpCodes focuses on providing digital building codes and regulations through its database, many of which were previously only available in physical print books. Copilot will pull from this database, which hosts over 5 million code sections, over 1,700 state and city codes, and 160,000 local amendments. With building codes often changing, UpCodes updates the searchable database about 7,000 times per month.
The free version of Copilot answers up to three user questions, while the professional and enterprise plans provide unlimited service. Users can refine searches by code years and building types.
The AI assistant can summarize and explain code sections, find related sections for research, perform calculations and create checklists based on project parameters. Example prompts are available to help users structure their questions, with the company suggesting they specify guidelines such as if the building is new or existing.
While the tool is capable of providing answers to code-based questions, it does require some fact-checking by users. To assist with this, the tool also cites relevant code sources so users can verify the building regulations themselves.
UpCodes says the Copilot assistant scored 74% on text-based code questions in the “Project Development and Documentation” section of the ARE 5.0 practice exam, 86% on the “B3 - Building Plans Examiner: Practice Exam” on Udemy, and 88% on the “B2 - Commercial Building Inspector: Practice Exam” on Udemy.
CEO Scott Reynolds told TechCrunch that UpCodes has contributed training data for AIs including ChatGPT and Google Bard. While this means its datasets are included in more popular AI services, they do not pull this data in real-time and thus are likely to include out-of-date codes.
To date, UpCodes has raised $7.6 million. The company says it has doubled team size, quadrupled revenue and grown its dataset from 2 million to more than 5 million building code sections since March 2021.
