The definition of a data product 🥜
There is no official definition of a data product (yet), but it can be described as follows.
A data product is a body of data in usable, understandable, and governed form, with a clear purpose and an owner. Its format varies with the use case. For an AI model it might be a data pipeline; for a business decision-maker, a report.
AI is hyped enormously, and for good reason — it is transforming how we work, make decisions, and automate processes. Many companies have already invested in AI solutions: building AI-assisted systems, experimenting with generative AI, and deploying machine learning models — and those investments will grow.
Expectations are high. It often feels as though something is missing. Why isn’t AI delivering on its promises yet? Why do so many AI projects stay at the experimental stage instead of scaling into real competitive advantage?
One big factor is data.
A data warehouse isn’t enough – AI benefits from governed data products
Too often, AI solutions are developed without the data they use being readily available, in understandable form, or even properly governed. This is where data products and data product management come in. They may not sound like the trendiest subject, but they are the key to scaling AI — and the deeper you go into the topic, the more fascinating it becomes.
Data products are not just a new data warehouse or one more system project. They are a way of organising and productising data so that it supports the business and makes effective use of AI possible. Well-defined data products make data discoverable, combinable, and governed — they are not just raw information in databases but clear entities with ownership, a governance model, and a clear purpose.
The subject is not discussed enough at business level, though. When AI comes up, the conversation often revolves around model accuracy, new AI features, and technical implementation, when the real challenge is most often in the underlying data. If AI is supposed to find and combine information quickly but the data is scattered and of poor quality, AI cannot produce value. According to McKinsey research, data quality problems can add as much as 20–30 per cent in extra costs to data-driven decision-making, because time goes into finding, correcting, and combining information (McKinsey, 2022).
According to Gartner, as many as 30 per cent of generative AI projects may be abandoned by 2025. The reasons are poor data quality, inadequate risk management controls, rising costs, and unclear business value (Gartner, 2024). That underlines the point that although massive sums are being invested in AI solutions, their success is not a given. Without governed data, AI solutions can remain merely interesting experiments that produce no business benefit.
How do data products solve the challenges of using AI?
Data product thinking and data product management answer many of the challenges of scaling AI directly.
When a company wants to build an AI strategy that produces business benefit, it is worth approaching the matter through data products.
Data products make data usable for AI – well-defined data products ensure AI gets the right, coherent, high-quality data, and that decisions are not made from arbitrary databases.
Data product management supports a horizontal technology stack – a technology stack dedicated to AI requires a well-governed data infrastructure that connects different systems and processes.
Data products make it easier to integrate AI into the business – AI must not remain a separate experiment; it has to work seamlessly within business processes.
AI observability rests on governed data products – AI must not be a black box; its decisions have to be traceable and explainable.
Managing data products reduces AI operating costs – according to Boston Consulting Group research, there is on average a 6–12 month lag between deploying AI solutions and the value they produce, because companies first have to solve problems of data quality and governance. When a data-product-based model is adopted, that lag can be halved (BCG, 2024).
It is important to recognise, though, that data products are not the beginning of everything — they come in at the right moment, once the direction and the need are clear.
Data products are not IT’s responsibility alone – they are a business competitiveness factor
Companies’ competitiveness does not come from AI technology alone, but from how well they combine their own data with AI — and how reliably and controllably that combination works.
That is why data products and their management should be discussed more at business level. The success of AI solutions is not IT’s responsibility alone — it is a strategic question that requires a new kind of organisational thinking about governing data and using it in AI.
How do you make sure your AI is built on reliable, governed data?
About the writer
Emma Hertzberg (LinkedIn) is a strategic, results-oriented designer with over 10 years of experience developing services and businesses. She specialises in systemic challenges and turns complex needs into practical, effective solutions — from processes and business plans to roadmaps, concepts, and strategies.
Hertzberg works as a Senior Strategic Designer at Solita, where design thinking, business focus, and top-level technology expertise come together. She is an insightful, future-oriented consultant who researches and experiments with the convergence of AI, data, and business, and its effects on organisations at a strategic level.
She has helped found and develop new business models and build the new amid disruption across sectors — law, design, and data-driven decision-making among them. Hertzberg also researched the interfaces between data science, design, and business in her master’s thesis, and brings strong practical consulting experience to that thinking.
Continue the conversation with the writer
I’d be glad to continue the conversation on subjects like:
· How can my company get started with data product thinking in practice?
· What tools and technologies are needed to manage data products?
· Which companies have succeeded in using data products to scale AI?
Enjoyed this article? Subscribe to get notified about new Tekoälyfoorumi articles straight to your inbox. We’d also hugely appreciate it if you left us a review on Google. Thank you. 💙
