The Real Test of Digital Policy: Can Your Systems Become an Operating System?
When I read a long policy review of China’s digital transformation agenda, one concern stands out: too many companies translate policy language straight into a project list.
Cloud migration, data platforms, AI pilots, security controls and new infrastructure all matter. None of them, by themselves, is transformation. The harder question is whether they help the company see the margin on an order sooner, anticipate a delivery failure earlier, stop departments from debating whose number is correct, and trace a failure back to the process owner who can fix it.
That is how I read the policy direction. The 2023 Overall Layout Plan for the Construction of Digital China frames digital development around foundations, integration, capabilities and enabling conditions. The data-institution agenda puts rights, circulation, value distribution and security in the same conversation. Together, they do not describe an IT procurement programme. They describe a company in which technology, rules and accountability must operate as one.
The familiar mistake: treating a management problem as an IT portfolio
In enterprise management, digital infrastructure can become a capital-expenditure line, data elements can become a request for a better dashboard, and security governance can become an access-control programme. Each action can be sensible. Together, they can still become an expensive, disconnected portfolio.
The missing question is commercial: which operating chain is worth redesigning first?
Take order-to-cash. If sales commitments, credit approval, available inventory, production scheduling, shipment, invoicing and collection sit in different systems and are interpreted by different owners, even an excellent data platform will not create a shared version of reality for one customer order. Reporting becomes faster; decisions do not necessarily become better.
By process redesign, I do not mean digitising the old approval form. I mean redefining how work moves, who makes a decision and who owns the outcome. Without that change, automation merely reproduces old friction at a higher speed.
Data governance is a question of authority before it is a question of data
Data governance is often framed as a warehouse, master-data or data-quality project. Those capabilities matter. Yet the hard part is usually in the meeting room: is revenue recognised at contract, shipment or collection? Who owns an inventory figure—the warehouse or planning? Who has the final say on a customer record?
These are not merely field-definition disputes. They are disputes over management authority. When metrics are inconsistent, the company does not first lack a BI dashboard; it lacks a mechanism for deciding which definition governs.
I would start with a handful of high-frequency decisions where a wrong call affects margin, delivery or cash. Assign a business owner, a data owner and a change-approval path to every critical metric. Technology should make data available, traceable and controlled. Business leaders must own the definition and the outcome.
This also makes security practical. Security is not a final approval gate; it means knowing from the beginning what data may be used, by whom, under which conditions, and how the organisation stops when something goes wrong.
From a management perspective, technology must serve operating choices
From an enterprise-management perspective, technology choices cannot be judged only by the design itself. What matters is translating commercial objectives into architectural constraints—and translating those constraints back into executable business trade-offs.
If the enterprise wants a shorter delivery cycle, the useful question is where the constraint actually sits: forecasting, scheduling, supplier coordination or exception response? Each answer calls for different data, interfaces and decision rights. If the enterprise wants lower inventory, a prediction model is not enough; inventory accuracy, replenishment authority and exception handling must also change.
The same logic applies to AI. I would not begin with “What can AI do?” I would ask, “Which decision chain needs to be shortened?” AI can retrieve, predict, summarise and manage exceptions. It cannot resolve vague authority boundaries or conflicting incentives. A small, accountable workflow with usable data and a verifiable result is usually a better starting point than a company-wide AI platform.
A management screen for digital investment
From a management perspective, I would ask four questions of any material digital programme:
- Which business outcome changes: revenue, margin, delivery, inventory, cash or risk exposure?
- In the redesigned process, who has decision rights and who owns the final result?
- Can the required data be traced to its source, definition, permissions and change history?
- Six months after go-live, what evidence will tell the enterprise whether to continue investing?
If these answers are absent, a project may still go live on time while the transformation fails. If they are clear, even a pilot around one order, one delivery flow or one region can develop into repeatable operating capability.
The policy lesson is not to create another layer of technology activity. It is to manage complexity more deliberately: allow leaders to see the same facts, make trade-offs faster, and place accountability where the work happens.
Sources
- Overall Layout Plan for the Construction of Digital China, State Council of the People’s Republic of China, 2023.
- Opinions on Building Basic Systems for Data, National Development and Reform Commission, 2022.
- Digital China Development Report (2024) release, National Data Administration, 2025.