When to use which chat
A decision guide for Ask, Focus, Co-work, and workflows: pick the right surface for the task.
docAnalyzer offers three chat modes (Ask, Focus, Co-work) plus a batch path called workflows. Each one fits a different shape of work.
Quick reference
| If you want to… | Open… |
|---|---|
| Explore your workspace, get oriented, plan the next step | Ask docAnalyzer |
| Get a deep, cited answer from a specific set of sources | Focus |
| Refine a note alongside the AI in a full editor | Co-work |
| Run the same task across every source in a set | a workflow |
When to open Ask docAnalyzer
Ask is best when the work isn't yet defined. You're asking "what's in here?", "where should I start?", "what does this product do?", or "I need to do X across these and those, how should we tackle it?". Ask will plan the steps and dispatch the right surface for each.
It's also the right place for product-orientation questions ("what file types do you support?", "how do credits work on Pro?"). Those answers come straight from product knowledge, not from your documents.
When to open a Focus
You're ready to dig in. You know which documents (or notes, or labels) matter. You want answers grounded in those sources with citations you can click back to.
Single-document, multi-document, and label-scoped chats are all just Focus. Pick the right dataset shape:
- One document: open Focus from the document's info panel.
- Several documents: multi-select on the documents page, then open Focus.
- A label: open Focus from the label; the dataset expands to every source the label holds.
- A mix: combine documents, notes, and labels in one dataset.
Datasets grow and shrink between turns. Add or remove sources without restarting the chat.
When to open Co-work
You have a note (written from scratch, saved from a Focus answer, imported as a markdown file) and you want the AI to refine it with you. Co-work edits the note directly: drafting sections, restructuring, rewriting tone, turning prose into a table. Every turn is small and reversible.
Co-work's context is the note itself. If you want answers grounded in your documents instead, that's a Focus.
When to run a workflow
If the task is "do the same thing across every document in this set", a workflow is faster and cleaner than a chat. Workflows are non-interactive and return structured output:
- Summarize a collection: one summary per document.
- Extract structured data: fields-to-spreadsheet across N documents.
- Audit against a reference: compliance report per document.
- Reduce AI signatures, Generate SEO metadata, Ask the same question of every document.
See When to run a workflow for the longer version of this decision.
Worked examples
"I just uploaded 30 contracts. What do I have?" → Ask docAnalyzer. It will list and characterize them, and probably suggest a workflow.
"Summarize each of these 30 contracts." → Workflow (Summarizer). Same task, every document.
"What are the renewal terms in the Acme contract?" → Focus on the Acme document. Cited answer that links to the relevant pages.
"Compare renewal terms across the 30 contracts." → Focus on the label that holds the 30 contracts. The dataset expands at retrieval time.
"Polish this note I drafted from the comparison." → Co-work on the note.