Have you ever considered the difference between a tool and a system?
Tools are what we use to perform tasks. Systems similarly streamline tasks.
Some might imagine a system as simply a more complex tool.
However, when tasks are categorized into two types—iteration work and flow work—the distinction between tools and systems becomes clear.
Iteration and Flow
Iteration work is the process of gradually creating deliverables through flexible trial and error.
In iteration work, a toolkit that can be used interchangeably for specific tasks is useful.
Conversely, flow work is a process that progresses incrementally, producing a deliverable at the final stage.
In flow work, having a system to guide the task flow significantly improves productivity and quality.
Flow Work Conversion and Systematization
Many tasks performed by humans are either iteration work or a part of systematized flow work.
By converting iteration work into flow work and then systematizing it, productivity and quality can be significantly enhanced.
The Industrial Revolution and the IT Revolution
The Industrial Revolution and the IT Revolution are prime examples of how converting iteration work into flow work, and then systematizing it, dramatically improved productivity and quality.
Prior to the Industrial Revolution, manufacturing was performed as iteration work, with humans skillfully using tools and freely altering arrangements and procedures each time.
Similarly, before the IT Revolution, information processing involved humans using tools in an unstructured, iterative manner.
By systematizing these processes, much like factory production lines or business IT systems, productivity and quality were significantly enhanced.
However, not only systematization but also the flow work conversion of iterative tasks is extremely crucial. It was the ability to convert to flow work that made systematization possible in the first place.
The Generative AI Revolution
When aiming to improve productivity and quality by utilizing generative AI in business operations, merely using AI as a tool will not unlock its true value.
The core objective is the flow work conversion of iteration work, followed by the systematization of that flow work.
Generative AI can handle iteration work due to its adaptability. However, whether performed by humans or generative AI, there are limits to the productivity and quality of iteration work.
Therefore, aiming for flow work conversion and systematization is crucial.
One might argue that if flow work conversion could improve productivity and quality for human workers, such initiatives could have been undertaken even before the advent of generative AI.
However, flow work conversion, when predicated on human workers, is actually a very difficult problem. Human workers cannot immediately adapt to changes in task assignments or content.
Conversely, when the worker is a generative AI, it is easy to iteratively reconfigure roles and task content.
Unlike humans, generative AI can forget previous steps, instantly read and understand new procedures, and perform tasks based on them.
For this reason, the mainstream approach to leveraging generative AI in business will be the flow work conversion of iteration work and its subsequent systematization.
Business Efficiency Using Generative AI
Let's consider an example of business efficiency achieved through generative AI.
As an example, consider the task of responding to employee inquiries about internal company rules.
Generative AI can be used to search internal rules and draft answers.
However, there is a possibility that generative AI might refer to outdated rules or mistakenly generate responses based on imagined information not present in the rules.
Furthermore, inquiries can come through various channels, such as email, messenger tools, phone, or in-person.
Therefore, the employee handling inquiries still needs to receive them as before.
It's conceivable that efficiency could be improved by having employees answer inquiries on the spot when possible, and for those requiring rule verification, inputting the inquiry content into generative AI to generate draft answers.
Additionally, for frequently asked questions, it's necessary to post them on the company's internal website as FAQs.
Generative AI can also be used to input representative questions and answers and generate bulleted drafts for website publication.
Moreover, generative AI can be leveraged to review draft wording when rule amendments are needed.
Such uses might streamline a certain percentage of inquiry handling tasks.
However, this merely treats inquiry handling as iteration work and uses generative AI as a tool.
Consequently, the efficiency gains from this approach are very limited.
Flow Work Conversion
To maximize the efficiency of the inquiry response task mentioned as an example, this task must be converted into flow work.
This requires detailing and documenting the steps taken by the person in charge when handling inquiries:
- Receive inquiries through various channels.
- If the inquiry is the same as one previously answered and there are no changes to the relevant rules, provide the same answer as before.
- For a new inquiry, or one involving a rule change, review the rules and prepare a draft response.
- Check if the draft response refers to old rules or includes information not stated in the rules.
- Check if approval is required before responding, and obtain approval if necessary.
- Respond via the channel through which the inquiry was received.
- Register the inquiry content, approval result, and response result in the inquiry history data.
- Periodically check the inquiry history data to create proposed updates for frequently asked questions and answers.
- Update the internal company website after obtaining approval.
- When rules are updated, update the rule data referenced.
- Concurrently, record in past inquiry history data that related responses and rule updates have occurred.
- Verify if frequently asked questions and answers need revision due to rule changes, and update if necessary.
By clarifying the details of these tasks, and connecting them, the flexible iteration work can be transformed into a clear flow work.
Example of Systematization
By converting tasks into flow work, the path to systematization becomes clear.
When systematizing, if some sacrifice in employee convenience is acceptable, one option is to consolidate inquiry channels.
Conversely, if employee convenience is the highest priority, all inquiry channels should remain open.
Fundamentally, the system should directly receive inquiries. Only in the case of verbal inquiries would a human input the details into the system.
After an inquiry is received, the IT system and generative AI will perform as much of the subsequent tasks as possible according to the flow. Initially, human checks and approvals should be interspersed throughout the system, and human operators should be able to make corrections.
Then, as the system is used to handle inquiries, if the generative AI makes a mistake, instructions for the AI should be updated with points of caution, items to check, examples of errors, and correct examples to prevent the mistake from recurring.
This process can reduce generative AI's errors. Updating these AI instructions itself can be made even more efficient by converting it from iteration work to flow work.
In this way, by systematizing flow-converted tasks, even operations that might initially seem to require human intervention can be replaced by a generative AI-centric system.
Common Misconceptions
Many people believe that the business application of generative AI currently has little effect, or that it is premature.
However, most of these individuals often have two types of misconceptions.
The first misconception stems from focusing on using generative AI as a mere tool.
As demonstrated here, leveraging generative AI as a tool for iteration work does not significantly boost business efficiency. This misconception arises from experiencing or observing such limited results.
The second misconception comes from focusing on having generative AI perform iteration work.
Indeed, attempting to have current generative AI perform iteration work is often unsuccessful. Consequently, people mistakenly conclude that generative AI cannot take over tasks performed by humans, based solely on this observation.
Conclusion
As discussed, by converting iteration work into flow work and systematizing it, greater efficiency can be expected than with tools alone.
Moreover, even if generative AI cannot perform iteration work, it can handle many individual tasks within a flow work process. Even if there are many errors initially, continuous improvement can be achieved by updating the instructions.
Alternatively, if necessary, tasks can be divided, separating drafting from checking, or implementing multi-stage checks.
If systematization can be achieved in this way, then improvements will progress with each task performed, and operations will become more efficient over time.
This is a method of working that allows for the continuous improvement of the mechanism itself, similar to factory production and IT systemization.
To effectively utilize generative AI, a shift in mindset is required: instead of trying to improve one's own iteration work, one must objectively convert one's tasks into flow work and systematize them.