As someone who’s led large-scale AI deployments at PETRONAS and spearheaded innovative initiatives at Astro, I’ve learned a thing or two about building and leading high-performing AI product teams. Today, I want to share my personal experiences and insights on how to assemble a world-class team, conquer new markets, and stay ahead in the ever-evolving world of AI. Buckle up – it’s going to be an exciting ride!

Assembling the Dream Team: More Than Just Technical Expertise

When I started building my AI team from scratch, I quickly realized that technical skills alone wouldn’t cut it. Here’s how I approached it:

1. Start with a Clear Vision

At PETRONAS, before diving into hiring, I worked closely not just with the CEO, but with a range of stakeholders to define our product vision. While I didn’t use the AWS Working Backwards method for this particular AI product, I had success with this approach when incubating a community-based parcel service. The key takeaway? Involving diverse perspectives in vision-setting is crucial, regardless of the specific method used.

2. Develop and Refine Your Total Cost of Ownership (TCO) Model

One crucial step that I’ve found invaluable, yet is often overlooked, is developing a comprehensive Total Cost of Ownership (TCO) model. This isn’t just a one-time task – it’s an ongoing process that’s critical to managing cost escalation risks and ensuring your solution remains competitive.

At PETRONAS, I led the TCO modeling for various applications we developed. This experience taught me the importance of continually refining our TCO model. Here’s why it matters:

1. Predictive Budgeting: A well-developed TCO model allows you to anticipate costs more accurately, helping you avoid budget overruns that can derail your program or platform.
2. Competitive Edge: By understanding your true costs, you can price your product more effectively and maintain a competitive edge in the market.
3. Risk Management: Regular refinement of your TCO model helps you identify potential cost escalation risks early, allowing you to take preventive measures.
4. Informed Decision Making: A detailed TCO model provides valuable insights that inform strategic decisions about product development, scaling, and market expansion.

In practice, I work closely with our Business Analyst, Product Manager, and Innovation Manager to define and regularly update our TCO model. We consider factors like infrastructure costs, data storage and processing, ongoing maintenance, and potential scaling needs.

One particularly effective strategy we implemented was setting up quarterly TCO review sessions. In these meetings, we’d examine our actual costs against our projections, identify any discrepancies, and adjust our model accordingly. This iterative process has been crucial in keeping our costs under control and our products competitive.

Remember, in the world of AI products, where costs can quickly spiral due to data and computational needs, a well-maintained TCO model isn’t just nice to have – it’s essential for long-term success.

3. Build a Cross-Functional Core

I learned that diversity is key. In my teams, I always ensure we have a mix of ML experts, software engineers, UX designers, and data scientists. But here’s the kicker – I look for people with cross-functional skills. For instance, I had a data engineer who was also proficient in backend software engineering. This dual expertise allowed them to think ahead and build data observability features directly into our platform. Similarly, our front-end developer had a solid grasp of UI/UX principles, effectively doubling as a UI/UX designer when working with project managers. These multi-dimensional skill sets proved invaluable in creating a more cohesive and efficient team.

4. Embrace AI-Assisted Tools (But Don’t Rely Too Heavily on Them)

In one of my projects, we used AI tools provided by platform to speed up our coding process. It was a game-changer! However, I always caution my team not to over-rely on external SaaS tools. Their roadmaps can change, leaving you in a lurch. Thus, we relied on Enterprise AI platforms and have a proper agreement. Startups as well as might can rely on AI-Assisted tools as long as it’s having no strong dependency on product features.

Table 1: Potential Roles, Skills and AI-assisted tools

Expanding Market Reach: Lessons from the Trenches

Conquering new markets with AI products is no walk in the park. Here’s what worked for me:

1. Do Your Homework

Before entering any new market, I always conduct thorough research. For instance, when I’m exploring opportunities, I discovered that the global digital avatar market has a staggering CAGR of 49.8%. If you see such numbers, you know you are entering into something big.

2. Collaborate with Client Innovation Teams

One strategy that worked wonders for me was partnering with innovation teams of potential clients. I remember countless coffee meetings with product managers, where we’d discuss their internal roadmaps and how our AI solution could fit in. These discussions were gold – they helped us tweak our MVP to match real market needs.

3. Navigate the Regulatory Maze

When we were launching an AI product in the APAC region, data privacy regulations were our biggest hurdle. Drawing from my experience at Astro, where I ensured ISO 27001 compliance for our data and AI platforms, I made sure our new product was compliant by design. It saved us a ton of headaches down the line.

Each industry and region has it’s own regulatory requirements. Here is the table which gives various regulatory framework to abide by when building an AI product. AI has to do with data and lot of data-related regulations kicks in.

https://claude.site/artifacts/260345d2-84db-467d-8782-dbeb424e7267

Staying Ahead of the Curve: Embracing Emerging Technologies

In the AI world, if you’re not moving forward, you’re falling behind. Here’s how I keep my team on the cutting edge:

1. Focus on Game-Changing Technologies

From my hands-on experience with Generative AI and LLM frameworks, I’ve seen how these technologies can transform products. In one project, we integrated Malay Language Model into our recommendation engine, and the improvement in user engagement was off the charts.

2. Foster a Culture of Continuous Learning

I’m a firm believer in learning by doing. In my teams, we regularly set aside time for experimenting with new AI technologies. It’s not always successful, but the learnings are invaluable.

Building a Resilient Team in the Age of Automation

As AI continues to automate many tasks, building a resilient team is more crucial than ever. Here’s my approach:

1. Encourage Multi-Dimensional Skill Sets

I always push my team members to step out of their comfort zones. For example, I once had a data scientist who showed interest in UX design. We supported her in taking on some UX projects, and she ended up bringing a unique data-driven perspective to our user interfaces.

2. Implement Smart Workflow Tools

I’m currently leading the development of AITaskBoard, a tool that automates much of the scrum master’s work. It’s all about working smarter, not harder.

3. Foster Open Communication

I believe in a relatively flat hierarchy. I make it a point to have one-on-one time with team members to understand their challenges and keep them motivated. It’s amazing what you can learn over a simple cup of coffee.

Wrapping Up: The Journey Never Ends

Building and leading an AI product team is a continuous journey of learning and adaptation. It’s challenging, but incredibly rewarding. Remember, the goal isn’t just to build a product – it’s to build a team that can continuously innovate and adapt in the face of rapid technological change.

As we stand on the brink of new AI frontiers – be it holographic interfaces or even more advanced generative models – I’m excited to see what the future holds. And I hope that by sharing my experiences, I’ve given you some insights to help you on your own AI journey.

Here’s to pushing the boundaries of what’s possible with AI. The future is ours to shape!