055 AI shop's Gemini CLI kills SaaS Enter full token into LLM
GeminiCLI can connect to the database and perform various operations simply by issuing a request in natural language.
This is a high-performance, high-priced, multi-functional SaaS database app service.
Its GUI screen
is difficult to customize, and
the usability is puzzling, whether it's a bug or a feature
. There are various opinions, but...
In other words, existing SaaS
can be used outside of fixed web applications.
Furthermore, multiple SaaS can be operated simultaneously.
In fact, the AI agent can simplify the database and
create new systems that achieve the main objectives.
[Goodbye GUI?] The day LLM will disrupt the conventional wisdom of SaaS development and utilization
"Where is this button?" "Which screen do I go to to use this function?"
Have you ever introduced a highly functional SaaS only to find yourself overwhelmed by its complex GUI and only being able to use a portion of the functions?
We are at a historic turning point that will fundamentally change how software is developed and used, and large-scale language models (LLMs) are at the center of it all.
In this article, we will explore the innovations in development methods brought about by LLM, the future of user interfaces, and new ways to use SaaS, based on our own experience of developing a simple CRM/ERP application from scratch using Gemini CLI .
- A new era of development style: "Creating apps through dialogue with AI" Traditionally, developing a business application has been a long process: defining requirements, creating thick design documents, selecting a framework, endless coding and testing... However, our development landscape was completely different.
What I did:
Gemini as a system architect : "I want to create a simple CRM with basic Salesforce and SAP functionality," I requested in natural language . Gemini then suggested a project name, a list of features, and the optimal technology stack.
Gemini as a programmer : I asked him to write a Python template based on the design. The main.pyframework and command skeletons were generated in an instant.
Gemini as a DB engineer : "Design the customer and product tables and implement the CRUD operation code." A consistent code set was completed, models.pyfrom the database model definition ( ) to the data manipulation functions ( ).crud.py
Gemini as a technical writer : I asked them to "create a README and operation manual for this app." Documentation for developers and users was automatically generated.
This is no longer traditional "development." It's "building applications through conversation" with a smart AI assistant . The prompts become the design document, the specifications, and the executable code itself. This unparalleled speed and flexibility cannot be compared to waterfall development or manual API implementation.
- LLM Interface Impact: Why GUIs are becoming obsolete The application we developed can be operated through a generated Python command line (CLI) or through natural language via Gemini CLI. Why not a shiny web interface (GUI)?
The answer is that GUIs are inherently "lame".
The GUI is nothing more than a combination of buttons and forms prepared by the developer, and cannot respond to slightly more complex ad-hoc requests such as "List only orders placed with Company A last month with a unit price of 100,000 yen or more."
The LLM interface, on the other hand, can be the ultimate all-purpose UI
: users communicate their requests in vague natural language, and the LLM interprets their intent and assembles and executes the necessary commands and database operations (SQL) on the backend.
[Future Concept: Dynamic Connector Development]
In this development, we had the LLM "create" a Python program that was then run by a human, but the next step is clear.
LLM directly interprets the DB schema and API specifications, and dynamically generates and executes "connector" programs for DB operations in real time according to user requests.
"Add a 'Customer Rank' column to the customer table and assign A, B, or C based on the total transaction amount over the past year."
By simply issuing these commands, LLM will be able to complete database migration, data analysis, and even updates. This future is just around the corner, and it will completely replace the old method (where humans read specifications and implement API clients).
- The Future of SaaS: From "Using" to "Using While Creating" The advent of LLMs will also revolutionize how SaaS is consumed.
① LLM as a "control tower" that integrates multiple SaaS
Many companies have implemented multiple SaaS services, such as Salesforce, SAP, Slack, and Microsoft 365, but data remains siloed within each service. While API integration is possible, development costs are enormous.
This is where LLM, the "universal translation machine and command center," comes into play.
"Get the business deals that have been 'won' in Salesforce, transfer the details to an SAP sales order, and register the expected delivery date in Google Calendar. Finally, notify all parties involved via Slack."
This single command in natural language will trigger the LLM to call the APIs of each SaaS in turn. SaaS may no longer be an individual application, but simply a single "function" that runs on the LLM operating system.
② A new paradigm: "Nurturing SaaS"
Traditionally, SaaS has been about using fixed functions provided by the vendor, but being able to build interactive applications is a different story.
The CRM we created this time only has basic functions right now, but if tomorrow you decide you want an invoice issuing function, you can simply ask Gemini to add a subcommand orderto the command create-invoiceand generate a PDF from the order information.
SaaS will change from something where you "buy a finished product" to something where you "develop functions on a daily basis through dialogue to suit your own business ."
Summary: The Dawn of a New Computing Platform
LLM is more than just a text-generation AI: it's a new computing platform that streamlines the software development process, redefines how humans and computers interact, and connects countless services .
This experience of developing a small CRM using Gemini CLI is just a glimpse into its vast possibilities.
Freed from the shackles of GUI, work progresses simply by thinking and interacting in natural language. Why not explore this new way of working and this new form of SaaS? A world of surprisingly simple and intelligent work awaits you.
table of contents
https://qiita.com/bono0/items/019f555e123b3f202279
Successfully develop the next generation original SaaS
055 AI shop's Gemini CLI kills SaaS Enter full token into LLM
Congratulations on the development,
-The End-
Top comments (0)