This article provides an overview of Conversational Search, an AI-powered way to search your Ironclad contracts using natural language. It combines AI-suggested filters with full semantic search across contract metadata and document text, so you can quickly find contracts and understand their terms.
Use Case
With Conversational Search, you can use natural language to surface relevant records and clauses, such as “NDAs with governing law California” or “MSAs expiring in the next 90 days”.
Prerequisites
| Features |
Dashboard |
| Permissions |
Admins must enable Conversational Search under Company Settings. Once enabled, this feature is available to all users in the instance, with results automatically limited by each user's existing permissions. |
Enable Conversational Search
Note: In-flight workflow documents are not included in search results.
To enable Conversational Search for your organization:
- In Ironclad, click your profile picture in the top right corner and select Company Settings from the dropdown menu.
- Click Ironclad AI from the left sidebar menu.
- Under the AI Features section, toggle on the Enable Conversational Search setting.
The primary search on your Dashboard will now use AI to suggest relevant answers.
Best Practices for Ironclad Conversational Search
Ironclad’s Conversational Search lets you use natural language to find contracts and understand their terms, without building complex filter sets or memorizing field names. These practices help you get fast, reliable results.
Ask naturally and be specific
Describe what you want in plain language. You can type a full question or a short natural-language phrase. You do not need special syntax to get started.
Prompts work best when they include, where relevant:
What you are looking for, such as an NDA, MSA, vendor agreement, or customer contract.
Key conditions, such as governing law, auto-renewal, or liability cap.
A timeframe or value range, such as next 90 days, signed in 2024, or above $250K.
A known property name or counterparty, such as Effective Date, Expiration Date, Counterparty, or Contract Value.
Example Prompt: Find NDAs with governing law California that expire in the next 12 months
Use the results list first, and use the AI summary when it appears
Search results remain the core output of Dashboard search. In record views, results can include in-context snippets so you can confirm why a contract matched without opening the file.
When Conversational Search invokes an AI summary or Assistant experience, use it for things like:
Counts, totals, or grouped answers across a result set.
Quick comparisons across matching contracts.
Follow-up prompts that refine the current result set, such as “Only show active contracts” or “Group by counterparty.”
Refine within the same search before starting over
A good pattern is to start with a focused prompt, then narrow the results with short follow-ups.
Example Prompt: Customer contracts with liability cap above $100,000 > Only those signed in 2024 > Limit to North America customers > Sort by expiration date
If an AI summary is available, follow-up prompts can apply to the current result set without restating the entire query.
Use filters when you need exact precision
Use natural-language search to get close quickly, then apply standard Dashboard filters to lock in the exact criteria you care about.
Once you have the slice of data you want, save the view so you can return to it later without rebuilding the search.
Know what Conversational Search searches today
For repository records and signed contracts, Conversational Search searches metadata and full document text.
For in-flight workflows, Conversational Search relies on metadata only. Unsigned in-flight workflow documents are not included in search results today.
Maintain data hygiene within your organization
Conversational Search can still help when metadata is incomplete or inconsistent, because it can use contract language and not just tagged fields.
Results are usually more consistent when:
- Key properties are normalized and maintained over time (e.g., record type, governing law, expiration date, auto‑renew, and contract value).
- You use tools like Smart Import and metadata extraction to bulk‑tag legacy contracts.
- You establish simple data hygiene guidelines for new workflows (e.g., required fields on launch).
For advanced connector-based search behavior, see Search the Dashboard.
Note:
Conversational Search respects the same workflow, record, and group permissions users already have in Ironclad. If someone cannot find a contract you expect them to see, confirm they have access to the relevant workflow or record and verify whether the content lives in the repository or in an in-flight workflow they cannot access.
Conversational Search FAQs
| What problems does it solve? |
Conversational Search helps users search in plain language instead of relying only on exact keywords or manually building complex filter sets. It also helps users validate search results faster with in-context previews and snippets. |
| What data does Conversational Search look at? |
Records and repository documents:
Workflows (in-flight contracts):
|
| How is Conversational Search different from Ironclad Assistant? |
Conversational Search lives in the search bar and is focused on finding relevant records, fields, and snippets from Dashboard search. Ironclad Assistant is the side-panel experience used for broader synthesis and next-step questions. A simple way to frame the difference is that Conversational Search helps with “where” and “what,” while Assistant helps with “so what” and “what next” questions. |
| Does Conversational Search work if our metadata is messy? | Yes, up to a point. Semantic search and full-text search can still help users find relevant contracts even when metadata is incomplete or inconsistent. |
| Does Conversational Search train on our contract data? | Conversational Search does not require model training on your specific contracts. It uses AI models at runtime, and separate AI training controls govern whether customer data contributes to broader model improvement across the platform. |
| How does it handle permissions and sensitive data? | Conversational Search respects the same permissions already configured in Ironclad. Users only see records and data they are authorized to access. |
Example Prompts
NDAs with governing law California that expire in the next 12 months
MSAs expiring in the next 90 days with auto-renewal terms
Vendor agreements with auto-renew in the next quarter above $250K in total value
Customer contracts with a liability cap above $100,000
Contracts that include a data processing addendum and mention sub-processors
All contracts with [Vendor Name] that have termination for convenience clauses
SaaS vendor contracts signed in 2024 with uptime SLAs below 99.9%
Customer agreements that allow unilateral price increases
Sales contracts closing this quarter with discounts greater than 20%
Contracts governed by UK law with arbitration seated in London