When global giants argue about which AI models are safest, smartest and ready for prime time, a New Zealander in Redmond, Washington is one of the people making the call.

Steve Sweetman leads the team at Microsoft's global headquarters deciding which AI models make it onto Microsoft Foundry – the platform that now offers more than 10,000 models from OpenAI, Anthropic, Meta, DeepSeek, Microsoft’s own labs and others, to customers around the world. 

Sweetman studied architecture at Unitec in Auckland and fell in love with technology as the industry shifted from manual drawing to computer-assisted design. Stints at Wang and Telecom led to Microsoft New Zealand, then on to Redmond, where he’s spent over two decades at the coalface of new tech: HoloLens, early chatbots, and the shift from narrow AI tools to today’s generative AI platforms.

Setting the responsible AI rules

He helped set up Microsoft’s Office of Responsible AI, turning high‑minded principles into practical policies and tools. That experience now shapes how Foundry works. Before any third‑party model is switched on for customers, Microsoft runs its own evaluations and has, Sweetman reveals, rejected popular models that didn’t meet its safety bar. 

Sweetman explains why the real value is no longer “just the model” but the data, governance and agent frameworks wrapped around it. We talk through concrete use cases from Alaska Airlines using generative AI to personalise travel planning, to CVS Health applying models to cancer research, and tiny Australian startup Lyrebird building multilingual tools to close gaps between patients and clinicians.

For New Zealand, where AI adoption is lagging and talent is thin on the ground, Sweetman is bullish. You don’t need to build your own foundational model, he argues. You can plug into powerful platforms like Foundry, understand the safety guardrails, and start experimenting.

If you want an insider’s view of how Microsoft is curating the world’s AI models – and a Kiwi’s take on what that means for local businesses – this is an episode worth queuing up. Streaming on iHeartRadio or wherever you get your podcasts. Thanks to our sponsor 2degrees and to Microsoft for hosting me in Redmond.

When global giants argue about which AI models are safest, smartest and ready for prime time, a New Zealander in Redmond, Washington is one of the people making the call.
Steve Sweetman leads the team at Microsoft's global headquarters deciding which AI models make it onto Microsoft Foundry &ndash; the platform that now offers more than 10,000 models from OpenAI, Anthropic, Meta, DeepSeek, Microsoft&rsquo;s own labs and others, to customers around the world.&nbsp;
Sweetman studied architecture at Unitec in Auckland and fell in love with technology as the industry shifted from manual drawing to computer-assisted design. Stints at Wang and Telecom led to Microsoft New Zealand, then on to Redmond, where he&rsquo;s spent over two decades at the coalface of new tech: HoloLens, early chatbots, and the shift from narrow AI tools to today&rsquo;s generative AI platforms.
Setting the responsible AI rules
He helped set up Microsoft&rsquo;s Office of Responsible AI, turning high‑minded principles into practical policies and tools. That experience now shapes how Foundry works. Before any third‑party model is switched on for customers, Microsoft runs its own evaluations and has, Sweetman reveals, rejected popular models that didn&rsquo;t meet its safety bar.&nbsp;
Sweetman explains why the real value is no longer &ldquo;just the model&rdquo; but the data, governance and agent frameworks wrapped around it. We talk through concrete use cases from Alaska Airlines using generative AI to personalise travel planning, to CVS Health applying models to cancer research, and tiny Australian startup Lyrebird building multilingual tools to close gaps between patients and clinicians.
For New Zealand, where AI adoption is lagging and talent is thin on the ground, Sweetman is bullish. You don&rsquo;t need to build your own foundational model, he argues. You can plug into powerful platforms like Foundry, understand the safety guardrails, and start experimenting.
If you want an insider&rsquo;s view of how Microsoft is curating the world&rsquo;s AI models &ndash; and a Kiwi&rsquo;s take on what that means for local businesses &ndash; this is an episode worth queuing up. Streaming on iHeartRadio or wherever you get your podcasts. Thanks to our sponsor 2degrees and to Microsoft for hosting me in Redmond.

The Business of Tech, hosted by leading tech journalist Peter Griffin. Every week they take a deep dive into emerging technology and news from the sector to help guide the important decisions all Business leaders make.

Issues such as cybersecurity, retaining trust after a cyberattack, business IT needs, purchasing SaaS tools and more.

New Episodes out every Thursday. Follow or subscribe to get it delivered straight to your favourite podcatcher.

@petergnz

@businessdesk_nz

Proudly sponsored by 2degrees Business!

 

The Business of Tech, hosted by leading tech journalist Peter Griffin. Every week they take a deep dive into emerging technology and news from the sector to help guide the important decisions all Business leaders make.
Issues such as cybersecurity, retaining trust after a cyberattack, business IT needs, purchasing SaaS tools and more.
New Episodes out every Thursday. Follow or subscribe to get it delivered straight to your favourite podcatcher.
@petergnz
@businessdesk_nz
Proudly sponsored by 2degrees Business!
&nbsp;

The Business of Tech

AI’s Kiwi gatekeeper inside Microsoft

https://api.omny.fm/orgs/7784f840-c291-422a-924b-ad9000bbad71/clips/0caf3db3-3b0d-42e8-8b92-b40a0017a0f4/transcript