Why Composer pauses to think
Simple questions get fast answers. But contract review isn’t simple. A request like “review this MSA for risk” requires Composer to:- Read the entire document
- Identify which provisions matter
- Check each one against your standards
- Decide what needs to change
- Generate redlines with explanations
What happens during thinking
Composer works through your request in stages: First, it breaks down the problem. A broad request becomes a series of specific tasks. “Review for risk” turns into checks on indemnification, liability caps, termination rights, IP ownership, and so on. Then it makes a plan. For research, this might mean formulating search queries. For contract review, it means deciding which clauses to examine and in what order. Then it executes. Composer works through each planned step: reading sections, applying playbook rules, comparing against your template, or searching external sources. Finally, it synthesizes. The findings from each step get combined into a coherent response with redlines, comments, or a summary.Status indicators
During thinking, you’ll see what Composer is doing:- Reading document
- Reviewing specific clauses
- Applying playbook
- Searching web sources
- Generating redlines
The benefit of step-by-step processing
This approach takes longer than a single-pass response, but produces better results:- Fewer errors because each step gets validated before the next one starts
- More thorough analysis because nothing gets skipped
- Clearer reasoning because the AI plans before it acts
- Better redlines because changes are informed by full document context
When to expect thinking
Composer uses extended thinking for:- Full contract reviews
- Multi-document comparisons
- Research requests
- Compliance checks
- Chained workflows (review → draft → summarize)