Two Keys for Small and Medium Enterprises to Leverage Generative AI

Small and medium businesses (SMBs) must focus on improving data quality and establishing an AI model adoption strategy to benefit from generative AI. Compared to large enterprises with greater resources, SMBs face the reality that AI implementation is not simply about purchasing tools—the foundational groundwork is equally critical. This perspective is becoming increasingly recognized.

Generative AI is a collective term for AI technologies that can automatically generate content such as text, images, and code. Over the past few years, practical implementation has accelerated rapidly, and its use has expanded primarily among large enterprises in areas such as marketing document creation, customer service automation, and operational efficiency improvements. Meanwhile, SMBs lag in adoption due to the costs associated with AI implementation and a shortage of specialized expertise.

In this context, the issue of "data quality" is frequently highlighted. Since generative AI learns from and references vast amounts of data to generate outputs, the reliability of the results decreases if the accuracy and consistency of input data are low. In many SMBs, customer information, inventory data, and past transaction records are not managed in a unified format, making it necessary to prioritize data organization before implementing AI.

Another challenge is the absence of a "model strategy." Generative AI models come in various types depending on their intended use and scale, and organizations must select one that aligns with their business operations, budget, and security requirements. The costs and benefits differ significantly depending on whether to use a general-purpose large-scale model as-is, deploy a specialized model for specific tasks, or conduct customization. Implementation without strategy carries the risk of reducing cost-effectiveness.

The high barriers to AI adoption for SMBs stem from a shortage of dedicated IT personnel and a cautious approach to initial investment. However, as cloud-based generative AI services have become available at low cost, the technical barriers to entry have been lowered compared to the past. Consequently, the actual barriers may lie more in defining the purpose—"what will we use AI for?"—and establishing the data infrastructure to support it.

Improving data quality and formulating a model strategy are not sufficient if pursued independently. Even with high-quality data, using a model misaligned with business objectives will yield limited results. Conversely, deploying an excellent model without organizing data will not produce accurate outputs. Treating these two factors as the two wheels of a vehicle and developing them simultaneously represents a realistic starting point for SMBs to apply generative AI to actual business operations.

Going forward, SMBs should focus on accumulating use cases tailored to industry-specific characteristics. For example, beginning with domains where impact is achievable despite smaller scale—such as inventory forecasting in retail or automated inquiry response in service industries—can lead to gradual expansion of AI utilization. Efforts to carefully combine data strategy with model selection hold the key to SMB success in AI implementation.