Hi, this is Nakanishi from the QA Group (though I also wear a few other hats at the Developer Relations Group and the KINTO FACTORY Development Group ^^)

This year at KINTO Technologies, we're embracing an "AI First, Release First" mindset. As part of this shift, our QA team has been exploring ways to make the most of AI to speed up our release cycles.

This time, a group of QA members who shared the same passion came together for a lively brainstorming session.

• "Wouldn't it be awesome if we could do this?"
• "I really want to try something like this!"

Together, we discussed ideas and possibilities.

In this article, I'll be sharing some of the most exciting concepts that emerged out of the session, exploring how AI can help transform the future of QA.

---

Issues That Emerged from Our Discussion

QA teams produce a huge volume of documents every day, but the sheer amount of information makes it hard to find what's actually needed. Specification formats also vary by project or person, which complicates sharing across teams. On top of that, there's no solid system for reviewing incidents or implementing preventive measures—making it tough to stop issues from recurring. The review process itself is also a heavy workload, and there's growing demand to streamline it.

A Future Made Possible by AI

Effective Information Use (RAG)

RAG (Retrieval-Augmented Generation) is a cutting-edge method for retrieving and analyzing information using the latest in generative AI technology. It quickly pulls key insights from vast archives of documents and incident data, delivering the right information to users when they need it. For instance, when an incident occurs, AI can instantly scan and analyze similar past cases to suggest effective solutions on the spot. It's like having a top-tier assistant who remembers every past experience and offers instant advice right when you need it. Already in use across industries like finance and customer support, it's dramatically speeding up response times and boosting problem-solving efficiency.

Organizing and Supporting Specifications and Designs

AI helps identify inconsistencies and omissions in your specifications, clearly highlighting and organizing any issues. By analyzing the inputted specifications, it can also automatically generate suitable test scenarios. For example, if you provide the specs for an e-commerce site cart feature, the AI can instantly create a scenario like: "Add product → Change quantity → Payment → Order confirmation." It can even generate additional scenarios, such as error handling flows and boundary value tests. This drastically cuts down the time and effort spent on manual scenario creation, boosting both the accuracy and efficiency of QA tasks.

Automate and Streamline Your QA Process

AI analyzes user operation logs to catch even the small mistakes that humans might miss. Specifically, it detects frequent user errors and unusual operations, like "errors triggered during specific screen transitions" or "fields often filled in incorrectly on input forms." With this, you can refine your test scenarios and address potential problem areas in advance. AI also automates the hassle of managing test case numbering on Conflu, cutting down on manual mistakes and saving a significant amount of time.

Incident Analysis and Prevention

AI that analyzes past incidents and offers concrete measures to prevent them from happening again. For example, it thoroughly reviews cases like "display issues on specific browsers" or "payment system failures" on e-commerce sites. Based on the findings, it suggests actionable steps such as "regularly checking browser behavior by version" or "enhancing error handling before and after payment processing." When a critical issue arises, the AI immediately assesses the risk level and sends automatic alerts to the relevant teams, enabling swift, real-time response and resolution.

Streamlining Test Data Creation

AI instantly generates the data required for testing. From new vehicle and used car details to user information, it can quickly produce large volumes of diverse data tailored to realistic business scenarios. In addition, by integrating AI with browser automation tools like Selenium and Appium, tasks that once required manual input can now be automated. With just a few simple settings, you can generate massive amounts of test data in no time. This integration not only reduces human error but also slashes the workload required for data creation, greatly improving the overall efficiency of your QA process.

Tool Integration and Process Automation

Connect Slack with tools like JIRA or Asana to receive timely, automated updates on what matters. Streamline your workflow by centralizing information from various tools in one place. Let AI serve as the bridge between platforms, smoothing out your entire process flow.

Using AI Models for QA

We're exploring the use of ChatGPT fine-tuned specifically for QA tasks. By building our own in-house AI model, we dramatically improved the response speed of the AI. Furthermore, we're working to create an environment where developers can easily use AI for quick self-checks.

---

Future Action Plans

1. Prioritize organizing and consolidating information using AI.
2. AI-assisted reviews will help us work more efficiently and lighten the human load.
3. Continuous data collection for the development of dedicated AI models.
4. Actively utilizing AI-based incident analysis for process improvement.
5. Automation between various tools will lead to further improvements in efficiency.

---

Even when something feels too big to handle alone, new paths can open up when we think together. Inspired by the ideas from this brainstorming session, we'll launch a range of QA initiatives powered by AI. If you're a QA engineer interested in exploring new possibilities with AI or taking on fresh challenges with us, we'd love to connect. Casual chats and info sessions are always welcome. Feel free to reach out!