What Tasks Does Generative AI Work Best For? How to Choose the Right “Areas” and “Methods” to Maximize Impact

Whether generative AI succeeds or fails is determined not by “tool selection,” but by “which business tasks (Where)” you apply it to and “how you use it (How).” It’s essential to clarify the outcomes you expect, the use cases, and how it will be used.

In this column, as a first step toward adopting generative AI, we explain how to build a strategy around “which business tasks to use generative AI for, and how to use it.”

Challenges Facing Japanese Companies

In Japan, many companies are struggling with labor shortages and operational inefficiencies.

Japan’s working-age population has been on a downward trend since peaking in 1995, and it is said that by 2050 it will fall to 55.4 million—26% fewer than in 2020. Japan’s productivity is also low, ranking 26th out of 38 OECD countries.

Generative AI is attracting attention as a potential “game changer” to address these challenges.

Benefits and Impact of Generative AI

The impact of adopting generative AI falls into three areas: improving operational efficiency, reducing costs, and increasing added value. “Increasing added value” means using generative AI to create and deliver new value that didn’t exist before—for example, improving the customer experience.

Compared with U.S. companies, Japanese companies have not made sufficient use of digital technologies to transform existing businesses or create new ones. Put differently, there is significant room to create new added value—such as new business opportunities—through the use of generative AI.

You can maximize the benefits of generative AI by looking not only at obvious, day-to-day pain points such as “efficiency” and “cost reduction,” but also at management-level challenges that drive added value, such as improving the customer experience.

What Is Generative AI?

Generative AI is an application that learns rules and patterns from vast amounts of data and can create original content based on them.

Generative AI works by predicting the next word based on the context it is given. Because it generates output by predicting the next word probabilistically, it does not “know” the correct answer. It is therefore very important to understand that, by design, generative AI’s responses may include errors.

How Can Generative AI Be Used in Business?

Application Areas: Which Business Functions Can It Be Used For?

The “application area” of generative AI refers to where in your business activities you use it. Depending on the application area, use cases broadly fall into two categories: “horizontal (enterprise-wide)” and “vertical (function-specific).”

“Horizontal” use involves rolling out general-purpose chatbots such as ChatGPT or enterprise tools such as Microsoft Copilot across the organization. 70% of U.S. Fortune 500 companies use Microsoft Copilot.
“Vertical” use embeds generative AI into specific business processes. While it can deliver direct economic impact, adoption remains limited.

By function, it has been noted that 75% of the value created by generative AI adoption is concentrated in four areas: “sales and marketing,” “customer operations,” “software engineering,” and “R&D.”

Going forward, it is strategically important to accelerate adoption of “vertical (function-specific)” generative AI, where tangible economic impact can be expected.

Concrete Tasks for Generative AI: How Can It Be Used in Day-to-Day Work?

“Concrete tasks” refers to how you want generative AI to work for you—in other words, how you use it. Ultimately, these uses can be grouped into two categories: knowledge acquisition and automation.

Knowledge acquisition means obtaining the various types of knowledge needed for work. Sometimes you need general knowledge, such as English vocabulary; other times you need highly specific, specialized knowledge, such as internal policies or product manuals. In many organizations, critical knowledge may reside “in someone’s head,” or it may take time and effort to find. Enabling people to quickly access the knowledge they truly need, when they need it—and having it presented in an easy-to-understand format such as summaries—is the first major benefit of using generative AI.

Automation means enabling tasks that previously required human judgment or flexible handling to be carried out automatically—for example, “creating meeting minutes and a to-do list from a recording,” or “classifying customer comments.” Even amid labor shortages, many processes are still run through inefficient, labor-intensive workarounds. By having AI take over those tasks, people can focus on higher value-added work that only humans can do. This is the second benefit of automation.

Because generative AI outputs may contain errors, regardless of whether you are using it for knowledge acquisition or automation, you need to design processes that use generative AI in the “upstream” step and have humans validate in the “downstream” step.

Rather than aiming for only one of these, many use cases seek to achieve both at the same time.

• Examples of work automation
  • Auto-generating job postings for HR: To reduce HR workloads, staff can automatically create job postings for the company simply by selecting relevant keywords.
• Examples of knowledge & insight acquisition
  • Financial market analysis and information services: A media company that provides financial information trains generative AI on its accumulated, massive financial datasets to deliver advanced market analysis and insights.
• Examples of work automation / knowledge & insight acquisition
  • Market research and campaign planning: For marketers, generative AI analyzes market trends, consumer reactions, and competitor activities to develop marketing plans.

Risks and Governance for Generative AI

The risks of using generative AI in business mainly fall into five categories: hallucinations and misjudgments, bias, information leakage, malicious manipulation of generative AI, and copyright infringement.

Among these, hallucinations and misjudgments occur frequently and require particular attention. AI works by predicting the next word based on the context it is given. It also tries to respond faithfully to the scope and content of the question. If the question itself contains assumptions or bias, the answer may be “pulled along” and become biased as well. When the person using generative AI has limited knowledge of the domain they are asking about, they may not notice bias or errors in the output. For example, whether it is translating from Japanese into English or writing code for software development, if specialists cannot review the results, embedding generative AI into that business process significantly increases risk.

From a management perspective, decisions are needed about “how much risk to take.” In addition, you will need operational rules that can be implemented in day-to-day work.

Analyze risks and determine—based on a management perspective—how much risk you should accept in light of the expected benefits of generative AI. Also consider how to manage those risks. Use the example below to analyze risks for your own organization.

Conclusion

Using generative AI effectively in business means identifying the business domain (Where) that ties to strategically important management challenges, designing how to use it (How), and operating it accordingly.