AI projects aren’t like traditional software projects. They require their own approach. For starters, you will need a strong data infrastructure to support them. But even more importantly, you need to think of AI development as an iterative process: creating, training, and deploying solutions in a production environment.
Before you can succeed, you need to know what success means. Success in AI projects is context-specific, but it usually starts with answering a few basic questions:
- What problem are we solving?
- How will we know when we’ve succeeded?
- What steps are needed to deploy this solution in production?
- What product, user, and infrastructure challenges do we need to address?
Once you’ve answered these questions, you can start thinking about the broader outcomes. What does success look like? Here are some possibilities:
- New Knowledge: Generating insights that weren’t available before, documented through reports or presentations.
- Better Decisions: Using those insights to inform policy or strategy.
- Product Features: Adding something like Gmail’s Smart Reply to an existing product.
- Standalone Products: Creating a data product or app that drives measurable business impact.
- Automation and Efficiency: Building systems that make processes faster, cheaper, or more accurate.
Each of these outcomes ties back to the same idea: delivering value. That’s what makes AI worth doing.
Building the Right Team
If you want to succeed with AI, you need the right team. What's the right team? the right team is diverse—people with skills that complement each other. Over time, you’ll want specialists: machine learning engineers, data engineers, data scientists, and AI product managers. But in the beginning, especially if you’re a startup, the same people will wear multiple hats.
Building this team isn’t just about hiring. It’s also about training. Organizations that invest in developing in-house talent will have a big advantage in the long run. They’ll be better equipped to handle the unique challenges of AI projects.
What does an AI team actually do? It depends on where you are in the process. Early on, the team’s job is to define an AI roadmap. This means looking at the next six months to two years and figuring out which emerging technologies are worth betting on.
From there, the team moves to execution. They’ll analyze business needs, build solutions, and set up the infrastructure to support them. This includes everything from collecting data to building prototypes to creating full-scale production systems. They’ll also be the ones translating cutting-edge research into tools that actually work.
One critical role of the team is to democratize data. Data can’t just be the domain of an analytics group or senior management. To really succeed, organizations need to make data broadly accessible (within legal limits, of course). That’s how you build a culture where everyone can make informed decisions.
In the beginning, the team you build might only be a couple of people. They’ll do everything: managing data pipelines, training models, building prototypes, and working with business units. But as the organization grows, roles will naturally become more specialized.
The key is to focus on delivering value from the start. Once you’ve done a few cross-functional projects, you’ll have a process in place for consistently delivering results. And at that point, scaling isn’t just easier—it’s inevitable.
In summary, AI projects can be challenging to pull off. Even when you’re working hard, it’s easy to get tripped up by the fast-changing technical landscape, confusion about how to organize teams, or the struggle to separate hype from reality. That’s why, before we dive into specific ML/AI projects, it’s crucial to take a step back and understand the challenges we’ll face in terms of the big picture. Once we have that perspective, we can start prioritizing ideas and evaluating their potential impact.
And in the long run, you’re aiming for something bigger—integrating AI into your workflows and building data products that capitalize on emerging technologies. That’s the exciting part of AI. Done right, it doesn’t just solve individual problems. It changes how organizations operate. It automates processes, creates new capabilities, and drives smarter decision-making.