The AI Workspace is the central panel of Error Analyzer, and it functions as far more than a static display of results. It is designed to act as a research copilot, an interactive partner that remains aware of your data, your analytical framework, and your immediate focus. By learning how to interact with it effectively, you move beyond simple output review and transform the AI into a responsive assistant that supports inquiry, annotation, and interpretation.
Part 1: What the AI Knows (Contextual Awareness) #
The Workspace is powerful because it remains aware of several layers of information simultaneously. Its primary focus is always the Error and Corrected text currently displayed on the screen. Alongside this, it integrates the linguistic blueprint produced by the NLP analysis, which means it has access to part-of-speech labels, grammatical structures, and named entities for every word and phrase. It also incorporates your tag library, understanding both the categories you have defined and the conceptual structure of your codebook. If you have already applied annotations, those are included as well, giving the AI a clear view of your ongoing analysis. Finally, when you highlight a segment of text, the Workspace narrows its scope to that specific unit. The highlighted portion is displayed above the chat as “Selected Text,” and any subsequent query is interpreted directly in relation to that selection. Together, these five layers of awareness ensure that the AI is not a generic assistant but a collaborator that adapts to the methodological context of your project.
Part 2: Your Controls (Tailoring the AI to the Task) #
You remain in control of how the AI behaves through two simple mechanisms. The first is model selection. The Flash model is optimized for speed and works best for straightforward checks such as identifying part-of-speech or listing verbs. The Pro model is optimized for depth, designed for tasks that require nuanced reasoning or theoretical interpretation, such as explaining why a learner might produce a particular error. A practical workflow is to begin with Flash for routine questions and switch to Pro when the analysis demands deeper insight.
The second mechanism is tool selection, accessible through the “+” button next to the chat input. In its default mode, the Workspace acts as a conversational partner, answering questions and clarifying structures. Selecting the Auto-Tag Tool turns the AI into an annotator, allowing you to issue explicit tagging commands. Choosing the Web Search Tool extends its reach beyond your dataset, enabling answers that integrate both project-specific context and information gathered from the wider web. Tools are deliberately active for one action at a time, and a small indicator confirms when a special mode is in use. This design prevents ambiguity and ensures that every action is intentional.
Part 3: Putting It All Together (A Practical Example) #
A typical workflow illustrates how these elements combine. Suppose you encounter the phrase he go in a learner text. You highlight the words, and the “Selected Text” indicator confirms the focus. With the Flash model active, you ask: Analyze the grammatical error in the selected text. The AI responds that this is a subject–verb agreement error and explains that the correct form is he goes. To turn this interpretation into structured data, you activate the Auto-Tag tool. A label appears confirming the mode, and you type: Apply the “S-V Agreement” tag to the selected text. The AI presents its plan: I will apply the tag “S-V Agreement” to “he go” and waits for your approval. Once you confirm, the annotation is applied, and a colored underline appears under the phrase. What began as a query has now become a verifiable annotation, directly linked to your codebook.
From Assistant to Copilot #
The AI Workspace combines contextual awareness with user controls to create a flexible environment for analysis. By guiding its model and tool selection, you ensure that the system operates at the right level of speed or depth for the task. By confirming every action, you retain methodological rigor while benefiting from computational efficiency. In this way, the Workspace evolves from being a passive feature into a genuine research copilot, one that accelerates routine annotation, strengthens consistency, and keeps your interpretive expertise at the center of the process.