🤖 Intelligent Assistants for IBM i: From a great ally to a new promise

For years, working with code on IBM i has been a task that requires experience, patience, and a lot of tacit knowledge. But thanks to artificial intelligence, today we have tools that accompany us, understand us, and even suggest how to improve our code.

💙 GitHub Copilot: A tireless companion

Ever since GitHub Copilot came into our lives, it became more than a tool: a silent yet powerful ally.

✅ It helped us write faster
✅ It completed code for us when ideas were scarce
✅ It recognized structures in RPG Full Free with surprising precision
✅ And it even suggested complex queries in SQLRPGLE that would normally take several minutes to build

🎯 Although it was not designed specifically for IBM i, it knew how to adapt, and many of us have adopted it as an essential part of our day to day.
🧠 We were surprised by its ability to understand patterns, adapt to the style of our code, and even help us train new programmers.

For all of that: Thank you, Copilot! 🙌

👀 watsonx Code Assistant for i: The promise that is approaching

Now, IBM goes one step further with an assistant designed specifically for IBM i:
IBM watsonx Code Assistant for i.

Although it is not yet generally available, what has been announced is very promising:

🔹 Reading and explanation of RPG code (Fixed and Free Format)
🔹 Support for legacy code modernization
🔹 Automatic generation of unit tests
🔹 Natural language conversations to understand business logic
🔹 Integration with VS Code and deployment in cloud/hybrid/on-premise environments

We are not talking only about code suggestions, we are talking about understanding the technical and business context of IBM i. Something that only an assistant trained with a focus on this platform can achieve.

🏗️ What do we know today about watsonx Code Assistant for i?

🧬 Modular and flexible architecture

Watsonx Code Assistant for i is built on the watsonx.ai platform, which uses IBM Granite foundation models, optimized for code generation and comprehension tasks.

It is composed of the following layers:

• Watsonx.ai foundation
• Watsonx Code Assistant runtime
• IDE extensions
• Connectivity with IBM i systems

Integration with tools such as ADDI for dependency analysis is also expected.

🔐 Security and privacy: pillars of the design

One of IBM’s strongest differentiators compared to other artificial intelligence solutions is its commitment to privacy and enterprise governance. That is why watsonx Code Assistant for i was designed with a focus on:

• Regulatory compliance: It complies with standards such as GDPR, HIPAA, and CCPA.
• Data security: It implements robust security controls, including:
  • Data encryption in transit and at rest
  • Use of customer-managed keys (BYOK)
  • Flexible deployment options
• ✅ Strict privacy policy: IBM does not use the content that customers upload nor the results generated by the foundation models to retrain or improve other models. All interaction with watsonx Code Assistant stays within the customer’s environment, ensuring the confidentiality, integrity, and ownership of the data.

🔧 Modernization paths in Watsonx Code Assistant: three complementary approaches

✅ 1. Automated Fixes

Watsonx Code Assistant detects obsolete or non-recommended code patterns and applies changes directly. This allows developers to focus on more complex and creative tasks, while the assistant takes care of routine improvements. For example, it can suggest replacing old code structures with new recommended practices, such as converting RPG Fixed Format code to Free Format. The idea is that the assistant makes changes that do not require manual review, improving code quality automatically.

🧭 2. Assisted Fixes

Proposals that the developer must review and approve before applying them. This allows the user to have the final control over the changes, ensuring that the modifications align with the specific needs of the project. For example, if the assistant suggests an optimization in an SQL query, the developer can review it and decide whether to implement it or not. This approach combines the intelligence of the assistant with the experience of the developer, enabling effective collaboration.

🛠️ 3. Self-Directed Fixes

The user directs the modifications with the support of the assistant as a copilot. In this case, the assistant acts as a guide, providing suggestions and resources while the developer makes the final decisions. For example, if a developer wants to refactor a complex function, they can ask the assistant to suggest best practices or code examples, but the final implementation remains in the hands of the user. This approach fosters developer autonomy and enables continuous learning.

🛠️ How to get high-quality code recommendations with Watsonx Code Assistant for i?

🧾 Best practices:

1. Use the iterative chat in the IDE
2. Reference specific files
3. Provide technical or functional context
4. Refine the result
5. Use natural language with technical terms

A good practice is to provide code examples or describe the problem context. For example:

Quiero que me ayudes a generar un programa en RPG Full Free que lea un archivo de clientes y muestre sus nombres y direcciones. El archivo se llama `CLIENTES` y tiene los campos `NOMBRE` y `DIRECCION`.

And if you need to generate base code, you can ask for something like:

Generame un boilerplate de un programa en SQLRPGLE que conecte a una base de datos y realice una consulta simple.

💡 A tip that is always important: Be clear when generating boilerplate code.

All of this helps the assistant better understand your needs and generate more relevant code. In addition, it is important to be clear and specific in your requests to get the best possible result. And it is very important not to forget that, although AI can do a lot, human judgment remains essential to validate and adapt the suggestions to your specific context. A good assistant does not replace the developer, but rather empowers them.

🧑‍💻 Supported IDEs

Watsonx Code Assistant for i will integrate with the following IDEs:

• Visual Studio Code: With a dedicated extension that will allow you to leverage all the assistant’s capabilities directly in the editor.
• RDi (Rational Developer for i): Integration for users who prefer this traditional development environment on IBM i.

This integration will facilitate access to the assistant’s features without needing to switch environments, allowing developers to work more efficiently. In the case of Visual Studio Code, the extension is expected to offer an experience similar to Copilot’s, but with a specific focus on the needs of IBM i developers. While for RDi, the integration will allow leveraging modernization and code generation capabilities directly in the most widely used development environment on IBM i.

🆚 Copilot vs watsonx Code Assistant for i

Now, let’s compare the capabilities of GitHub Copilot and watsonx Code Assistant for i:

While GitHub Copilot has been a great ally, watsonx Code Assistant for i promises to take the development experience on IBM i to a new level, with a deeper focus on the specific needs of this platform. Although Copilot has proven to be a valuable companion, the specialization of watsonx Code Assistant in the IBM i ecosystem could make a significant difference.

🧭 Final reflection

GitHub Copilot earned our trust, it is a great ally for IBM i developers and its ability to adapt to our needs has been impressive, and its impact on productivity is undeniable. With this, Watsonx Code Assistant for i has a high bar to clear, but it is already showing clear signs that it can do so, and if it delivers on its promise, it will be a turning point for IBM i. With the different intelligent assistants that are emerging, such as watsonx Code Assistant for i or GitHub Copilot, we are facing a new era of development on IBM i, where artificial intelligence not only accompanies us, but also understands us and enhances our capabilities, allowing us to focus on what really matters: solving problems and creating value for our users, but it is very important to remember that AI does not replace the developer, but empowers them. We must remain critical and careful with the suggestions we receive, always validating their relevance and quality, ensuring that they align with our needs and quality standards. This is key to making responsible and effective use of these tools, ensuring that artificial intelligence is an ally on our path toward the modernization and continuous improvement of our systems.