The application period for the Small and Medium Enterprise Agency’s “Digitalization and AI Introduction Subsidy” has begun. According to ITmedia Business Online, key changes from the previous system include the establishment of a new category specifically for AI utilization and revisions to subsidy rates and upper limits. This news is likely drawing the attention of many executives and DX promotion managers, building momentum to apply.
However, let’s pause and think for a moment. This subsidy is ultimately support for the “means.” What should truly be questioned is whether the executive themselves can define the “purpose”—the “why” behind the AI introduction. The tailwind of a subsidy risks accelerating IT investment without purpose, potentially even accumulating “digital debt.”
This article uses this latest subsidy news as a starting point to explain the trap of “lack of an objective function” that executives must absolutely avoid when introducing AI, and what specifically should be defined.
The Fundamental Question Hidden by the Subsidy
The new subsidy system is certainly attractive. A dedicated category for AI has been created, making it easier for more companies to take on the challenge. This is undoubtedly a tailwind. However, from a management x IT perspective, there is a major pitfall here. It’s the tendency for the thinking of “we’ll introduce it because there’s a subsidy” to take root, where the means become the end.
What we have witnessed through supporting over 38 clients is the scene of tools and systems being introduced for reasons like “because we got a budget” or “because it’s trendy,” resulting in a scattering of “unused assets” within the company. SaaS subscription costs pile up, and no one can grasp the overall picture. This is a typical result of management defining the “purpose” of IT, focusing only on selecting “means” and executing the budget.
AI is more powerful than any previous IT tool. Therefore, the potential “harm” from introducing it without a purpose can also be greater. Automating business workflows without a plan merely replicates wrong decisions at high speed. If an AI for management analysis proposes optimization based on disparate metrics for each department, the company’s overall direction will only become more fragmented.
Before creating the subsidy application documents, the question executives themselves should ask is this single point: “What is the ‘objective function’ for our company’s AI introduction?”
Defining the “Objective Function”: The 3 Management Layers
An “objective function” is a mathematically expressed goal that instructs AI or IT systems on “what to maximize and what to minimize.” In management terms, it’s about translating “our company’s definition of success” into a form that IT can understand. If this is vague, even if the AI functions, it won’t serve management’s needs.
Executives need to consider and verbalize the objective function across the following three layers.
Layer 1: The “Speed” and “Quality” of Business Growth
If using AI for sales increase or customer acquisition, what is the objective? Is it to “maximize the number of inquiries”? Or to “optimize the discovery of high-conversion leads”? Or perhaps “customer service that considers Customer Lifetime Value (LTV)”?
For example, when introducing AI features into a marketing automation tool (like HubSpot), automating “sending as many emails as possible” versus automating “learning from past response rate data to maximize engagement by identifying optimal timing and phrasing” will yield completely different results for the business. The latter requires defining an objective function like “engagement.”
The executive’s responsibility is to move beyond “I want to increase sales with AI” to a level of specificity like: “I want the AI to maximize the accuracy of discovering potential customers with similar profiles within our existing customer data by learning the characteristics of our high-value products.”
Layer 2: The “Reproducibility” and “Speed” of Management Decisions
This is an area many executives overlook, yet it is extremely important. AI has the potential to bring reproducibility and speed to management decisions that previously relied on experience and intuition.
For example, digitizing invoices and receipts (using tools like Money Forward Cloud Expense) is not just about operational efficiency. By combining AI with this, beyond automatic expense categorization and fraud detection, one might define the purpose as “automatically analyzing the time-series changes in the cost structure of each project and providing early warnings of signs of declining profitability.”
The objective function there would be to “maximize the ‘predictive accuracy’ of project budget vs. actual variance.” This shifts management decisions from a feeling of “expenses are high this month” to ones based on reproducible data like: “Based on patterns from similar past projects, Project A has an 85% risk of profit margin decreasing by 10% in three months.”
Layer 3: Liberation from the “Stability” of Task Execution
The evaluation of many IT departments tends to be tied to “stable system operation (zero failures).” While this is important, if this alone becomes the objective, it can become a factor inhibiting change and risk-taking. Setting the purpose of AI-driven task automation (think of it as the next generation of RPA) solely to “reduce human error to near zero” is a tremendous waste of potential.
For instance, when introducing an AI chatbot for internal help desk operations, the purpose could be defined as: “To reduce the workload of first-line response by 100%, shifting IT staff resources to more strategic security measures or to new SaaS selection projects through collaboration with business departments.”
The objective function is not “minimizing response time,” but “maximizing the proportion of human resources redistributed from routine tasks to strategic tasks.” This lays the foundation for transforming the IT department from a cost center into a value-creating department that supports management decision-making.
Practical Pre-Application Checklist: Specifying the Objective Function
Before starting to write the subsidy application documents, discuss and document the following items with your management team.
- Identifying the Problem to Solve: Be specific. Not “work is slow,” but “the manual invoice processing performed by 3 accounting staff members working overtime for the last 3 days of every month.”
- Definition of Success (KGI): What constitutes success for that problem post-implementation? Set quantifiable metrics, e.g., “reduce processing time by 80% and cut human resources by half.”
- Decision Criteria Required from AI (Objective Function): To achieve the above success, what kind of “judgment” do you want the AI to make? E.g., “Automatically categorize invoice items based on past classification patterns with a misclassification rate of 1% or less.”
- Data Verification: Does sufficient quality and quantity of past data exist to train the AI for that judgment?
- Redefining the Human Role: After the AI handles that part, what will the freed-up human resources be deployed to? Does the value of that new work exceed the cost of AI introduction?
This process is precisely the act of executives not “avoiding” IT (here, AI) and defining its purpose. The “Business Plan” and “Expected Effects” sections of the subsidy application should be filled with the output of this very discussion.
Conclusion: The Subsidy is Fuel to Accelerate a “Pre-Defined Purpose”
The Small and Medium Enterprise Agency’s Digitalization and AI Introduction Subsidy is powerful support for Japanese companies to enhance their competitiveness through technology. However, it is merely “fuel.” Deciding the “destination” and the most efficient “route” is the responsibility of the executives themselves.
Before installing the powerful engine that is AI, draw the blueprint for the vehicle (the organization) and input the precise destination into the navigation system (the objective function). This fundamental step tends to be neglected amidst excessive expectations for technology and the tailwind of subsidies.
Please view this system revision not merely as an opportunity for IT procurement, but as a chance to redefine “how to engage with AI from a management perspective.” Implementation without definition will inevitably rebound on the company in the form of future “digital debt,” even if the actual burden is reduced by the subsidy.
Why not start a strategic discussion by first facing a whiteboard and asking: “What is the objective function for our company’s AI introduction?”
Comments