Predicting Hidden Demand with AI – What Data and Structured Processes Teach About Scaling Complex Business
What if instead of forecasting what customers ask for, you started from the need itself? At our Data & AI breakfast event, Kesko’s Senior Data Scientist Marko Laakso explored how AI can uncover hidden demand and turn it into sharper operational decisions and clearer priorities.
Hidden demand doesn't show up cleanly in data. It surfaces as anomalies, recurring deviations, situations where something doesn't quite add up. People often sense first that something is off even without being able to name what. That intuition is a signal of some need that is unmet.
AI's role here is to help turn that intuition into something concrete. To identify what has changed. To model processes across a wide set of available resources. To find what no single observer can see alone.
Business Logic Is the Only Way to Manage Complexity
Interpreting data is just the beginning. Organisations constantly run into the same problem: individual solutions can be improved and adjusted locally, but the impact is often limited. Local optimisation isn't enough when the challenge is systemic.
Laakso raised something essential: to make significant changes, you need the ability to renew multiple areas simultaneously. That demands a shared understanding of the current situation. This forms the basis for all transformation. It requires a common language across the organisation, not just within data teams or leadership.
In practice, that means making business logic visible:
What are we building?
Who is responsible?
How do different solutions depend on each other?
Without answers to these questions, AI initiatives remain isolated experiments that never scale. And as Laakso noted, the lower the transition expenses, the more realistic it becomes that changes can actually be applied.
Machine-Readable Rules: The Bridge Between Business Logic and Implementation
This approach involves business modeling and building a semantic layer – a structured way of expressing business logic as machine-readable rules that binds together concepts, constraints, data sources, regulation and the actual code implementation.
The same definitions used within the documentation are used within the code itself. This ensures that AI assistants operate within the same rules as the people working alongside them. Change something in the code, and the documentation reflects it. Change a business rule, and the implementation follows.
The underlying point is that AI-assisted development only works if it has something to lean on. Without structured business logic, an LLM generates code in a vacuum. Its power lies precisely in using context. And that context has to be built first.
What Does This Mean in the AI Era?
In many organisations the gap between idea and implementation is still wide, even as AI lowers the technical barrier. LLMs are good at coding, testing, naming and structuring. But framing business problems, managing dependencies and anticipating the consequences of change all require people with a clear structure that AI can work within.
Laakso also highlighted the pace of development. We cannot solve everything at once or rush with technology. As he put it: "we should surf on that wave rather than going before it."
The smarter move is to build the semantic layer, the business logic, the shared language so that as the tools mature, you're ready to use them well.
Hidden demand is found in data. But responding to it requires an organisation capable of acting in a coordinated way within a complex environment. A tool won't get you there. A structure will.
Marko Laakso, Kesko
Senior Data Scientist
Marko Laakso applies AI and data science to large-scale customer and business data. His background spans biomedical research, software architecture and analytics – over two decades of turning complex data into decisions that matter.