Orica sets up AI governance to scale

Orica has stood up a “single front door” to keep track of AI use cases and their associated risk assessments, development and value realisation.

Image credit: Orica.

The mining explosives giant said last year that it is managing its AI demand and governance from within ServiceNow; the structure it is using for that is what is known as the ‘AI control tower’.

The vendor describes the control tower as “a centralised governance and oversight layer … designed to provide transparency, control, and assurance over AI models and agents running within ServiceNow.”

Orica’s senior manager of AI portfolio and delivery Irene Klymenko told the recent ServiceNow AI Summit Melbourne that use cases had been “popping up” across the company’s operations, with governance and progress tracked via spreadsheets.

“Our AI risk assessments were done on spreadsheets, which were emailed to peers and stored in a SharePoint folder somewhere, making it really difficult to track changes, owners, actions, mitigations or progress,” Klymenko said.

Orica worked with Deloitte to stand up the AI control tower, which serves three key purposes.

First, it is used to collect all AI use case proposals, to prioritise and categorise them, and to conduct and store risk assessments, which can be referenced at any point in the AI’s development and implementation.

Klymenko said that use cases are put forward both by internal staff as well as by vendors.

“It became quite important for us to have that visibility across what was happening in the organisation, in the workplace,” Klymenko said.

Second, the control tower also enables the use case development activity itself to be centrally monitored, and third, for tracking of whether an investment in AI has realised its anticipated value.

Klymenko said that the organisation had embraced the structure provided by the AI control tower.

“We were met with not resistance, but actually demand, as teams really saw this as a way to streamline and fast track their use cases,” she said.

“AI governance isn't meant to block you or to slow you down. It's actually the opposite. We want to get you away from spreadsheets into a place where we can have the process streamlined and make it a lot faster for you to use AI in the workplace.”

Although eighteen months into having the AI control tower, Klymenko indicated there are still some improvements to be made.

One of these is tighter integrations between the AI control tower and hyperscale cloud platforms, where some AI agents and tools are run.

“Integrations, as with any tool, [are] very challenging,” Klymenko said.

“Because we work across a number of hyperscalers, that's going to be something that we will continue to work on to make sure they're all built into our AI control tower.”

She also said some enhancements would be made in the way AI value is measured “through financial metrics”.