Every company has messy data, and even the best of AI companies
Every company has messy data, and even the best of AI companies are not fully satisfied with their data. If you have data, it is probably a good idea to get an AI team to have a look at it and give feedback. This can develop into a positive feedback loop for both the IT and AI teams in any company.
Hear now the words of Andrew Ng, a teacher of machines and of men, who declared: “Every company has messy data, and even the best of AI companies are not fully satisfied with their data. If you have data, it is probably a good idea to get an AI team to have a look at it and give feedback. This can develop into a positive feedback loop for both the IT and AI teams in any company.” At first these words seem technical, bound to the realm of numbers and machines. But within them lies a wisdom much older: that imperfection is not a barrier to progress, but the very seed from which growth springs. For messy data, like life itself, is raw, unrefined, and yet filled with possibility.
The origin of this saying rests in Ng’s long experience at the forefront of artificial intelligence. He has seen the grand laboratories of industry, the best-equipped research groups, and he knows their secret: none are ever fully satisfied with the purity of their data. There is always error, always noise, always incompleteness. Yet this is no cause for despair. For in these imperfections, AI can still learn, adapt, and reveal truths. His counsel is simple: do not wait for perfection before beginning, but let collaboration and feedback between IT and AI transform weakness into strength, disorder into positive loops of improvement.
History offers us a mirror of this truth. Consider the great voyages of discovery: when Columbus set sail, his maps were incomplete, his knowledge flawed, and his understanding of the world deeply imperfect. Yet he embarked, and through his voyages new continents were revealed to the world. The maps were messy data, filled with gaps and errors, but in using them, explorers refined them, and the maps grew ever more accurate. In the same way, Ng teaches us that waiting for perfect data is folly; it is through use, feedback, and continual refinement that knowledge matures.
The meaning of his words is also a lesson in humility. No company, no matter how advanced, can say its data is flawless. This confession strips away the illusion of perfection and calls us instead to the discipline of iteration. To bring IT and AI together, to allow one to critique and improve the other, is to enter into a cycle where imperfection breeds progress. It is the same cycle seen in all great endeavors—science, art, philosophy—where error and correction, failure and feedback, forge paths toward mastery.
But beyond humility lies hope. For Ng speaks of the positive feedback loop, a phrase of immense power. It means that improvement breeds further improvement, that small gains can cascade into great transformations. Just as a seed grows into a tree and a tree into a forest, so the sharing of feedback between human teams and artificial minds can create growth beyond what either could achieve alone. The messiness of data becomes not a curse but the soil in which innovation takes root.
The lesson for us, whether we build companies or live ordinary lives, is this: do not fear imperfection. Do not delay action while waiting for perfect conditions. Begin with what you have, let others give feedback, and allow improvement to grow step by step. Your “messy data” may be your incomplete plans, your flawed beginnings, your uncertain efforts—but if you share them, refine them, and return again with courage, they will become the foundation of greatness.
Practically, this means inviting collaboration into all things. Do not hoard your work in silence, but let others examine it, critique it, and shape it with you. Whether in business, in art, in family, or in learning, welcome feedback. See mistakes not as shame but as teachers. Just as AI thrives not on perfect data but on iterative refinement, so too does the human spirit thrive when it chooses humility, openness, and persistence.
So let us take to heart the wisdom of Andrew Ng: “Every company has messy data… but through feedback, it can become a positive loop.” Let this truth remind us that greatness does not rise from perfection but from persistence, not from flawless beginnings but from constant growth. And let us live in such a way that our messiness is not hidden, but transformed—into wisdom, into progress, into enduring light for generations to come.
NHToan Nguyen Huu
Reading this, I feel motivated to explore the intersection of data quality and team collaboration. What kinds of feedback mechanisms are most effective for fostering a positive loop between IT and AI teams? Are there tools or platforms that facilitate this process efficiently? It also raises the question of how to measure the impact of improved data quality on AI model performance and overall business outcomes.
PLTran Phuong Linh
This quote emphasizes that no dataset is ever perfect, even for leading AI companies. I wonder if Andrew Ng recommends specific methodologies for auditing or cleaning data before analysis. Could encouraging this type of proactive collaboration help companies identify new use cases for existing data? It also sparks a broader discussion on organizational culture—how openness to feedback and experimentation can make AI initiatives more successful.
SVCty TNHH Viet Sao Vang
I find this perspective practical and strategic. However, it raises questions about the human resource and time investment required to maintain such a feedback loop. Can companies create scalable systems that allow AI and IT teams to continuously improve data without overburdening either side? It also makes me think about how this collaboration could foster innovation and cross-team learning beyond just data quality improvements.
HPhan phan
This statement makes me reflect on the iterative nature of AI development. Could engaging AI teams to review data also help uncover hidden patterns or biases in datasets? It prompts consideration of how messy data not only impacts AI performance but also highlights organizational inefficiencies. I’m also curious about the best practices for establishing communication channels that facilitate continuous feedback between IT and AI teams.
HNHoai Nguyen
Reading this, I feel curious about the practical implementation of a positive feedback loop between IT and AI teams. How often should data reviews occur, and what metrics indicate improvement in data quality? I also wonder if smaller companies with limited AI resources can realistically benefit from such collaboration, or if this advice mainly applies to organizations with mature AI infrastructure.