Using AI in SAS Viya Workbench
SAS Viya Workbench is fundamentally a coding environment. Users spend much of their time writing, debugging, documenting, testing and maintaining code. Increasingly, AI is becoming a powerful tool for accelerating each of these activities. Whether you want AI Assistance or you want AI Delegation, Viya Workbench has options for how you can incorporate AI into your workflows.
Assistance vs Delegation
Before we get into exactly how we can use AI within Viya Workbench, let's first just clarify what we mean by AI Assistance and AI Delegation.
In an AI Assistance model, the human remains at the center of it all. Examples of AI Assistance include things like code completion, refactoring suggestions, test and document generation, and explaining code. We’re using AI to accelerate development, help the human with a specific, well-scoped task “explain these lines of code”, “add comments to this macro”, but the AI is not driving.
In an AI Delegation model, the human describes the outcome and “delegates” the implementation and execution to the AI. Examples of AI delegation might be telling an AI to “fix the build issues” or “create an analysis workflow”. Here we’re perhaps expecting our AI agent to write one or more programs from scratch. The point being that the AI performs multi-step work with minimal human intervention during the process. The AI is using its reasoning and ability to call different tools and services to achieve the human-defined outcome.
Rather than competing approaches, assistance and delegation exist on a spectrum. As confidence, context and task complexity increase, users may choose to move from AI helping with individual actions to AI completing larger units of work independently.
In Viya Workbench, these two patterns map naturally into two different ways of working with AI. IDE extensions provide AI assistance directly within your editor, helping you write, understand, and improve code while keeping you in control. CLI-based agents take the next step, enabling AI Delegation, where you can assign larger outcomes and allow the agent to perform multi-step work on your behalf. Understanding this distinction helps you choose the right tool for the task.
IDE Extensions
IDE extensions are the most accessible way to adopt an AI assistance workflow within SAS Viya Workbench. In this model, the human remains firmly in control while AI helps accelerate common development tasks. Rather than asking the AI to solve an entire problem independently, you are asking it to help with a specific piece of work: explaining code, generating comments, suggesting refactoring opportunities. For many users, this is the ideal starting point for incorporating AI into their daily workflow. The feedback is immediate, the interaction feels natural, and you can review every suggestion before deciding whether to accept it.
This approach maps naturally to the way most people already work within Viya Workbench. As you write code, AI can suggest completions. When working with an unfamiliar codebase, it can explain macros, DATA steps, PROC calls, or even entire programs. When preparing work for colleagues, it can help generate comments and documentation. The AI acts as a knowledgeable assistant sitting alongside you, helping you move faster while leaving all key decisions in your hands.
Since GitHub Copilot capabilities are now deeply integrated into the Visual Studio Code experience, users will find AI assistance available immediately. Other extensions can be installed directly from the Extensions tab within VS Code. Many of the major AI providers offer extensions that integrate directly into the editor (OpenAI’s Codex, and Anthropic’s Claude Code). Because extensions are installed into your Viya Workbench home directory, they persist across Viya Workbench sessions, meaning they only need to be installed once and will remain available whenever you launch a new Viya Workbench environment.
Users who prefer notebooks – don’t worry! AI-assisted development is increasingly available within Jupyter notebooks and JupyterLab environments, providing a similar experience.
The greatest strength of IDE-based assistance is that it keeps the developer in the loop. This makes AI assistance particularly well suited to learning new techniques or understanding unfamiliar code. However, there are times when the work extends beyond the scope of a single file or immediate coding task. You may want to implement a feature that spans multiple components, investigate issues across an entire project, or automate a sequence of development activities. In these situations, the objective often shifts from "help me do this" to "do this for me". This is where AI delegation and CLI-based agents become a natural next step.
CLI Tools
While IDE extensions are focused on AI assistance, Command Line Interfaces (CLIs) enable AI Delegation.
In an assistance workflow, we ask AI to help us complete a task. In a delegation workflow, we describe an outcome and allow the AI to determine how best to achieve it. We might ask an AI agent to investigate a dataset, integrate with a new data source, generate documentation, or improve the structure of an existing project. This shift from "help me do this" to "do this for me" is one of the most significant changes that AI is bringing to software development.
As with IDE extensions, all of the major AI providers now offer CLI-based tooling. Viya Workbench currently does not include npm which is the most common cross-platform way of installing these tools. However, all of the providers support a shell-based installation approach (GitHub Copilot, OpenAI Codex, Anthropic Claude Code). In the same way as the IDE extensions, the default is that these tools are installed into your home directory meaning that they persist across Viya Workbench sessions and only need to be installed once.
The ability to delegate larger tasks can significantly accelerate development, particularly when working with unfamiliar codebases. However, delegation also introduces a different responsibility model. The broader the scope of the work being delegated, the more important it becomes to review the results. Code, documentation, configuration changes, and generated assets should be validated with the same level of scrutiny that would be applied to work produced by a human colleague.
AI agents can be remarkably effective, but accountability for the final result remains with the person using them. The goal is not to replace human judgement, but to allow humans to spend less time on implementation details and more time defining objectives, evaluating outcomes, and solving the business problems that matter.
Conclusion
AI in Viya Workbench is not a single capability. It is a collection of workflows that lets you choose the right level of AI involvement for the task at hand. IDE extensions provide AI assistance, helping you write, understand, document, test, and improve code while keeping you firmly in control. CLI-based agents extend this model through AI delegation, allowing you to describe an outcome and have an agent complete broader, multi-step work on your behalf. For many users, assistance is the best place to start. As your confidence grows, and as your use cases become more complex, delegation can become a natural next step.
By supporting both approaches, Viya Workbench enables teams to adopt AI at a level that matches their skills, workflows, and appetite for automation. The key is not to use the most autonomous tool available. It is to choose the right level of AI involvement for the work you need to accomplish.
Call for content now open!
It's your turn to help shape SAS Innovate 2027. Share your expertise and inspire the SAS community.
