Skip to main content
Composer is an agentic system. When you give it a complex request, it breaks the task into sequential steps and executes them one at a time. The output from each step informs the next, creating a chain of reasoning that produces more accurate results.

​
Why this matters

When you ask “review and redline this contract,” Composer doesn’t try to do everything at once. Instead, it:
  1. Reads the full document
  2. Identifies the relevant issues
  3. Prioritizes by risk or importance
  4. Generates redlines for each issue
  5. Adds comments explaining the changes
Each step builds on the previous one. The AI validates its own work between steps, catching errors and ensuring consistency. This produces better redlines, fewer mistakes, and clearer rationale than a single-pass approach. Drafts stream in real time as the AI writes, so you see the language it’s generated.
Draft Interrogatory

​
The two-step pattern

The most effective way to prompt Composer is to separate thinking from acting:
  1. First: Review (think)
  2. Second: Revise (act)
Example: 
Step 1: Review this contract and identify all issues, prioritizing by risk level.
Step 2: Revise the contract to resolve the identified issues with redlines and comments.
By explicitly separating these steps, you give Composer full context before it makes changes. The result is more accurate redlines and comments that reflect deeper understanding.

​
Using verbs to guide workflows

The verbs you use in your prompts determine what kind of step Composer executes. Choose verbs intentionally to build effective workflows. Analysis verbs (thinking steps):
  • Review: Examine and evaluate
  • Identify: Find specific items
  • Compare: Analyze differences between documents
  • Flag: Call out specific issues
  • Prioritize: Rank by importance or risk
Action verbs (execution steps):
  • Draft: Create new language
  • Revise: Modify existing language
  • Redline: Generate tracked changes
  • Summarize: Condense information
  • Extract: Pull out specific data
Example workflow using verb progression: 
1. Review this MSA for liability issues (analysis)
2. Identify the three highest-risk provisions (analysis)
3. Draft fallback language for each (action)
4. Revise the fallbacks to match our standard tone (action)
5. Summarize the changes for the business team (action)
Each verb triggers a different type of processing. Analysis verbs make Composer think and evaluate. Action verbs make it produce output. Sequencing them properly creates reliable, repeatable workflows.

​
Workflow examples

​
Comprehensive contract review

Review incoming third-party paper against your standards.
  1. Upload the contract and your template
  2. “Compare this contract to our template and identify all deviations”
  3. “Which deviations pose material risk to us as the customer?”
  4. “Redline the contract to address the material risks with comments”

​
Learning from prior redlines

Apply consistent positions based on how you’ve handled similar contracts.
  1. Upload the current contract plus 2-3 prior redlined versions
  2. “Review how we’ve handled similar contracts. What positions did we take on indemnification, liability, and termination?”
  3. “Apply those positions to this contract with redlines”

​
Compliance review

Check a contract against specific regulations or policies.
  1. Upload the contract and relevant reference materials (regulations, checklists, policies)
  2. “Review this contract for compliance with [HIPAA/GDPR/etc.] and flag any gaps”
  3. “Suggest specific language to address each gap”
  4. “Draft redlines incorporating the compliant language”

​
Template comparison

Identify where counterparty paper deviates from your standard form.
  1. Upload both documents: their version and your template
  2. “Compare clause by clause and create a table of material differences”
  3. “Which differences favor the counterparty?”
  4. “Draft redlines to bring key provisions in line with our template”

​
Composer vs playbooks

​
Use playbooks when

  • You have recurring contract types with consistent criteria
  • You want standardized, repeatable outputs
  • Speed is critical for high-volume work

​
Use Composer when

  • Dealing with novel or one-off contracts
  • Handling redlines not covered by your playbook
  • Conducting legal research
  • Need flexibility in review approach

​
Use both when

  • Run the playbook first for standard issues
  • Use Composer to address remaining items
  • Generate summaries after playbook review

​
Effective prompting

DoDon’t
Use action verbs (review, revise, identify, draft)Use vague language (“make this better”)
Structure multi-step requests explicitlyCram everything into one unclear request
Provide context about your position (vendor/customer)Assume AI knows your business context
Specify desired output format (redlines, summary, list)Leave output format ambiguous
Include relevant documents as attachmentsReference documents the AI can’t access

​
Sample prompts

Contract review: 
Review this vendor agreement as the customer. Identify provisions 
that pose high risk, focusing on liability caps, termination rights, 
and data security. Then revise those provisions to be more favorable to us.
Targeted revision: 
[Highlight the limitation of liability clause]
Revise this section to:
- Exclude indirect and consequential damages
- Cap direct damages at $500K or 12 months of fees
- Carve out breaches of confidentiality from the cap
Issue spotting: 
Review this MSA and create a prioritized list of:
- Deal-breaker issues (things we cannot accept)
- High-priority negotiation points
- Minor issues to raise if time permits

Then prepare redlines to address the top 5 issues.

​
Troubleshooting

AI makes changes in wrong location Highlight the specific section before prompting. AI provides incomplete redlines Request explicitly: “Review the entire contract from Section 1 through the end.” Too many style changes Specify: “Only suggest substantive changes, not stylistic improvements.” Missing company context Add information to your Company Profile so it’s automatically included.