There is a growing need to add AI capabilities to existing web and business systems.
For example, you want to assist with an inquiry in AI chat, you want to automatically determine the image, you want to summarize the sentences entered, you want to analyze the accumulated data, and so on.
However, AI functions are not stand-alone, and how they connect to current systems and workflows is important.
The first thing I want to check is the current system configuration. Organize which languages and frameworks are built, what databases are used, whether there are integrations with external APIs, and what login and permission management looks like.
Next, clarify the tasks you want to delegate to the AI.
If you just want to include AI, the development scope will be too wide. Depending on whether the contact is a primary response, a classification of the image, a summary of the data, or an input aid, the required mechanism will change.
It is also important whether people check the results of AI or automate processing. Depending on the operation, it may be safer for a person to confirm the final form rather than confirming the judgment of the AI as it is. To be honest, AI cannot deny that there is still instability.
Experience has shown that adding AI to an existing system is less likely to fail if you start with a small feature than if you build it big from scratch.
To get started, it's a good idea to start with areas that are easy to see, such as contact handling, writing, image checking, and data organization.
Makoto Tejima supports everything from checking existing systems to designing and developing AI functions. Even if the content is not solid, you can organize it from the current issue.
