• Patrick Leung first worked at Google in 2007.
• He was blown away when he first saw a demo for Google’s AI assistant, Duplex, and wanted to join the team.
• Leung said his top tip for landing AI jobs is getting real-world experience.

This as-told-to essay is based on a conversation with 53-year-old Patrick Leung, a former Google employee who lives in California and works as a CTO. The following has been edited for length and clarity.

In the early 2000s, when I was living in New York City and running my own SaaS company, an early Google employee moved into my building.

We became friends quickly: he was from New Zealand, like me, and a brilliant, super charismatic engineer. I thought that if Google was full of people like him, I should consider working there.

I applied to Google and landed a job as a manager on the team of the online payments system, Google Checkout, in 2007, based in New York.

It was incredible to work at Google and not just because of the free food, famous speakers, and crazy office space. There was an atmosphere of revolution, like we were there to make real change.

I spent over 10 years there, and in early 2017, I pivoted to working on Google Duplex, a calling assistant that uses AI. At that time, it felt like AI was starting to take off, and we were at the forefront.

I wanted to experience working in different parts of Google

I spent a couple of years working on Google Checkout, before taking a brief break from the company to work for a startup in Palo Alto. I returned to Google in December 2009 and spent time on several teams, including Google Maps, emerging markets, and corporate engineering.

One day, I asked a colleague who’d recently moved to a new team what she was working on, but she said she couldn’t really talk about it. It was really unusual for people to be tight-lipped about their work, so I was curious.

I had no idea what I was walking into, but Google Duplex blew me away

I spoke to her manager, who showed me an early demo of the project, which was called Blue Ginger before it became Google Duplex.

It was an AI calling up a restaurant and booking a reservation. Even at that early stage, the voice sounded so real that it reminded me of the movie “Her.” I was blown away.

I wanted to join the team. They were only accepting internal transfers at that point: they had to know who you were and what you’d done. A part of me really liked that exclusivity.

I had a bunch of conversations with the team manager so he could decide if I was a good match. I highlighted my previous career accomplishments and ability to quickly contribute ideas. I can only speculate, but I’d say he chose me for a combination of my communication skills and my track record at Google — I’d launched other products in the past, which probably helped establish my credibility.

I had to retool myself to work on an AI-focused team

I began managing a team working on a visual interface to help human operators assess and correct how the Duplex AI system interpreted phone conversations, to improve the model.

I’d been exposed to machine learning and AI concepts on my previous teams at Google, but I’d never been on one actively building models and doing core data science work.

I brought my experience building software and managing people to the team, and while some members had more expertise in AI, most of us had to retool ourselves to learn this stuff.

I set up long discussions with others on the team who could explain how the system worked, but I wish I’d spent more time meeting with the other AI teams at Google, to get a feel for how they applied AI and machine learning.

At big companies, there are teams of extraordinary people all over the place. If I’d taken advantage of that environment more, I could have met with people who formed Google Brain or DeepMind and learned from them while we were at the same company and were able to be more open with me about their work.

When we launched Google Duplex, we knew it was going to be a big deal

In the lead-up to the launch, we wondered how the product would be received because of growing caution about AI taking people’s jobs.

There was a media frenzy when it launched in 2018. Some people thought it was unethical, and others thought it might be fake. Stephen Colbert even did a satirical skit about us. It was a privilege to be part of a launch that touched the zeitgeist of popular culture.

After 10 years with Google, the time came to move on. In 2019, I joined a financial sciences company as their CTO for private investments. Currently, I’m a CTO at a company called Faro Health, where I’m using AI to improve clinical drug trials.

My top tip for breaking into AI is to build up real-world experience.

If it’s possible in your current job, seek out opportunities to apply large language models to a real business problem. For example, a friend of mine with no coding experience used AI to personalize outreach messages while recruiting for clinical positions, which improved her response rates. If there aren’t opportunities like this at work, do it in your spare time and put it on your résumé.

If you’re looking to work in a certain area, it would probably help if you built a system that was directly applicable to it, but it could even just be something you’re interested in, like summarizing movie reviews.

The barrier to getting involved in this field is lower than it’s ever been. You no longer have to train your own model; you can just start talking to the large language model and do interesting things. You don’t need to be an expert in machine learning to make things happen. Even coding is becomingautomated thanks to vibe coding.

People who demonstrate that they can wield LLMs effectively are going to find jobs.

Google did not immediately respond to a request for comment from Business Insider.

Do you have a story to share about pivoting into AI? Contact this reporter at ccheong@businessinsider.com.