Pairing geotechnical data with AI helps New Zealand build better

Christchurch quake

The NZGD was born in the aftermath of the devastating earthquake that struck Christchurch in February 2011. A total of 185 people were killed, thousands displaced and much of the city center's infrastructure was badly damaged.

Engineers involved in rebuilding efforts urgently needed data on sub-surface conditions to make critical decisions, such as whether a structure could be safely repaired or restored, or needed to be demolished entirely.

Ellis-Garland recalled working on around 4,000 assessments for homes that needed to be rebuilt. "It's sensitive work when people are displaced and so we had to do our work really quickly," she said.

But collating that data was a challenge.

"It can be pretty fragmented," said Beca CEO Amelia Linzey.

She explained that results of underground scans or tests are collected by different agencies or private developers when a site is being built. Being unable to access such data meant engineers had to spend time and money drilling fresh boreholes to collect new data or, frustratingly, the same data that someone else might already have.

Creating a common repository to centralize and share all that information was the obvious solution.

"That was really the sentiment behind the NZGD," said Linzey. "Getting that data together as a national asset."

The NZGD was established in 2013 by the Canterbury Earthquake Recovery Authority. Ten years later, the Ministry of Business, Innovation and Employment invited proposals to host and upgrade the platform. By then, the NZGD had amassed thousands of users and around 168,000 geotechnical tests had been uploaded and shared.

NZGD 2.0, on Microsoft Azure

Beca submitted its idea for the next version of NZGD to use BEYON, its digital twin platform. This idea was accepted, making Beca the next custodian of the NZGD.

In November 2024, it launched the updated NZGD, which now runs on an SQL database on Microsoft Azure and is accessed via BEYON.

"Our focus was making it modern, secure, standards-compliant and integrated with modern spatial analytics technologies," said Beca's Witherden. "So that allows us to have better data quality, usability and access."

Beca opted for digital twin technology, said Witherden, because it's the best way to curate and manage complex, interconnected engineering information.

Shifting NZGD's technology stack to Azure made it more accessible, scalable and secure, said Witherden, with access controls like Azure Entra ID. It also set the foundation for more bells and whistles.

In late 2025, Beca added an agentic AI layer in BEYON, enabling over 4,300 NZGD users to filter, query and extract geotechnical data using natural language. The AI assistant was designed in Microsoft Foundry, with strict rules guiding its responses.

"Using Microsoft Foundry has been such a great tool for us because it has built-in guardrails, and that allows us to control how the AI responds," said Witherden. "It's not allowed to do geotechnical analysis."

Having all this data and technology on tap means users like Ellis-Garland can make better decisions, faster.

The AI assistant not only zeroes in on the investigation logs for her; it helps her cut through the ensuing information overload.