What does winning a hackathon tell us about successfully implementing AI solutions for real business value?

A few weeks ago, my team and I participated in Maersk’s AI Hackathon—an intensive, cross-functional event designed to encourage rapid innovation. Emerging as winners was thrilling, but more importantly, it provided a compelling blueprint for how to implement generative AI solutions that deliver tangible business impact. This experience wasn’t just about code; it was about choosing the right problem, collaborating seamlessly with the business, leveraging the best tools, and setting the stage for scalable growth and transformation.

Context: Making Complex, Unstructured Data Accessible

Our specific use case focused on Maersk’s Supplier Review team, who assess a wide range of suppliers against rigorous sustainability, health & safety, and compliance standards—including the prevention of slavery and child labour. Their challenge was daunting: suppliers submit supporting documents in many formats—often scanned PDFs or images—written in multiple foreign languages (for instance, scanned Chinese character documents). Processing this information manually was slow, labour-intensive, and prone to delays.

We set out to reimagine this workflow. Our goal was simple yet ambitious: make complex, unstructured data accessible to AI-driven business processes. By employing a combination of OCR, text extraction, translation services, and vector databases, we transformed scattered, hard-to-parse documents into a structured knowledge resource. This shift empowered AI tools, like chatbots and eventually multi-agent analysis systems, to provide clear, actionable insights—turning a task that once took over two days into something achievable in under 30 minutes.

1. Start with a High-Value, Real-World Use Case

In our hackathon, we didn’t begin by picking a fancy algorithm. We started with a real, pressing problem: The Supplier Review team needed a faster, more consistent way to evaluate supplier documentation for compliance. This wasn’t just about making a process more efficient; it was about fundamentally changing how insights are derived, enabling quicker, more thorough assessments that help maintain Maersk’s ethical and sustainability standards.

Takeaway: Find a problem that matters. Early successes—like reducing a two-day review process to 30 minutes—demonstrate value, build stakeholder confidence, and pave the way for broader AI adoption.

2. Use the Right Tools for Rapid Innovation and Scale: DIFY as a Game-Changer

Time is precious in a hackathon setting—and in business. We leveraged DIFY, a powerful orchestration framework, to rapidly develop and integrate our end-to-end solution. With DIFY, we could easily connect to Azure Blob Storage, incorporate OCR and translation services, and feed all extracted information into a vector database. This approach let us focus on creating value rather than reinventing infrastructure.

DIFY’s flexible integration model and ready-to-use tools helped us stand up a proof-of-concept quickly. As we move beyond the hackathon, DIFY’s orchestration and pipeline capabilities set the stage for further refinements—introducing multi-agent process flows, custom code optimizations, and seamless scaling. We’re confident we can evolve from a simple chatbot interface answering supplier queries to a more complex workflow that auto-generates compliance reports and flags potential issues proactively.

Takeaway: Start fast with platforms like DIFY that are both powerful and easy to integrate. Establish a CI/CD pipeline to maintain quality and future-proof the solution. Balancing speed and scalability ensures that early wins don’t become future headaches.

3. Foster a Culture of Collaboration and Experimentation

This success story wasn’t crafted by developers alone. We worked closely with the Supplier Review team to understand their challenges and priorities. Their input shaped our data pipelines, validation steps, and the final user interface. In turn, the technology team’s insights helped them see new possibilities—like leveraging generative AI not just to automate, but to reimagine how assessments are performed.

In a production environment, fostering this kind of dialogue between business and tech teams is essential. Embrace a culture where everyone is encouraged to raise questions, experiment with new ideas, and learn from quick iterations. This collaborative environment allows organizations to align their AI initiatives with genuine business objectives, leading to better outcomes and stronger stakeholder buy-in.

Takeaway: Open, cross-functional collaboration accelerates learning and innovation. It ensures that AI solutions fit into real-world contexts and directly address the pain points people face.

4. Think Beyond Automation—Reimagine the Process

At the heart of this project lies generative AI’s greatest promise: the ability not just to streamline existing processes, but to transform them. By converting complex, multilingual, and unstructured documents into accessible knowledge, we didn’t just make the old process faster—we enabled entirely new capabilities. Going forward, we envision a multi-agent AI system that can automatically assess compliance, suggest areas of concern, and even recommend best practices, going far beyond mere data extraction.

This shift is about more than efficiency; it’s about pioneering new ways of working. Generative AI can supply actionable insights, spark innovation, and unlock value streams that traditional approaches never even considered.

Takeaway: Use generative AI as a catalyst for reinvention. Don’t just settle for doing the same thing faster—think about what entirely new processes or business models become possible.

5. Embrace Continuous Learning and Evolution

Winning the hackathon was a milestone, not a finish line. AI technology, tools, and best practices evolve rapidly, and organizations need to evolve with them. Maintain a mindset of continuous improvement—iterate as you learn, update your models and pipelines as better services emerge, and encourage teams to share knowledge internally and externally.

In our case, future iterations might include more robust language capabilities, deeper integrations with internal data systems, or advanced reasoning agents that handle edge cases automatically. By committing to ongoing learning, we ensure we remain at the forefront of what’s possible.

Takeaway: Stay curious and nimble. Each project is a step forward in an ongoing journey, not a static achievement.

In Summary: Our success in the Maersk AI Hackathon offers more than bragging rights. It showcases a roadmap for how to implement AI—from selecting a meaningful use case to leveraging robust platforms like DIFY, fostering collaboration, reimagining business processes, and committing to continuous improvement.

Whether you’re an executive evaluating strategic investments or a developer coding tomorrow’s solutions, these lessons apply universally. Start practical, think big, adapt quickly, and never stop learning. That’s the formula that took our solution from a concept to a real, transformative business capability—and it can guide your AI journey as well.