Zak Kostura of Arup argues that digital tools pay off most when clients build the same facility many times. He says firms should build AI workflows off the project, stay problem-driven rather than data-driven, and treat software as a new front door.

Zak Kostura, Associate Principal, Digital Services Portfolio Leader for Americas East, Arup

Watch the episode ↗

In this conversation

How a documentary about the Citicorp building led Zak to structural engineering, and how designing Mexico City's airport roof meant writing software to design the building.

Why Arup splits AI into using third-party products and building its own models, and develops its own AI workflows off the project until they are mature.

Where AI earns its place in design: when speed matters, such as testing wind comfort, solar or biodiversity across many options.

Why owners of data centers, chip fabs and warehouses who build the same facility many times let a consultancy take a product approach to design.

How a modular chip-fab client got one design that can be checked against any site in the world, backed by a graph database of swappable parts.

His argument that digital tools can become a firm's new front door, bringing clients to the consultant through a different channel.

We've learned that trying to develop your own AI workflows from scratch is not compatible with the pace of a project.

Zak Kostura

Chapters

Designing software to design buildings

Zak's path from a Citicorp documentary to structural engineering and the Mexico City airport roof, and then to leading Arup's digital team.

Technologists plus consultants

What his team does day to day, and how the project phase and the design team's priorities narrow the technology choice.

Building AI off the project

Using AI products versus developing models, and why speed is the reason to bring AI into design workflows.

Consistency makes products possible

Why repeat-build owners enable a product approach, and how Arup incubates software close to the people who use it.

One design for any site

The modular chip-fab project, the site suitability tools, and why good solutions start from the problem, not the data.

A career like a jungle gym

His advice to young engineers to embrace change, and why the designer's role will not go away.

In order to develop good digital solutions you really have to be problem driven. You don't start with the data.

Zak Kostura

Takeaways

Developing AI workflows from scratch does not fit the pace of a project, so Arup builds them off-project and brings them in once they mature.

AI earns its place in design when speed matters, such as rapidly testing wind comfort, solar or biodiversity across many options.

Owners of data centers, chip fabs and warehouses who repeat one design let a consultancy take a product approach to design.

Good digital solutions start from the problem, not the data; for a modular chip-fab client that led to a graph database of swappable design parts.

Incubating products inside a firm keeps development close to users, though truly scaling a product can require outside capital.

The tools, the methodologies, the workflows and the role of you as a practitioner are going to change decade by decade, so be open to changing yourself.

Zak Kostura