Analysis Settings - AI Assistant
What this feature is
Analysis Settings is where you configure analysis fields: values your AI generates after a conversation ends. Those values can be attached to the person’s contact profile or to the conversation itself depending on how you set each field up. You describe what to extract or compute using rules (natural-language instructions). Rules can be reused across assistants so you define logic once instead of copying the same instructions into every assistant.
If you used Replify before this upgrade, this page replaces the older “Custom Summary Instructions”, “Custom Qualification Instructions”, and “Custom Next Step Instructions” experience. See How this relates to the old Analysis Settings below.
Video walkthrough
This narrated tutorial walks through Analysis Settings step by step. Your browser does not support embedded video.
Why it’s valuable
- One rule, many assistants: Set an org-wide default or share override logic where it makes sense, instead of duplicating long prompts on each assistant.
- Structured outputs: Fields have names, types, and update behavior so outputs stay consistent for your team and downstream tools.
- Automations and CRMs: Custom fields can flow through your Zapier integration so you can branch workflows, update records, or run post-processing outside Replify.
How this relates to the old Analysis Settings
Previously, Analysis Settings focused on three global custom instruction blocks for your assistant (qualification, conversation summary, and next steps). That worked, but the instructions lived per assistant and were easy to repeat or drift out of sync across a large team of assistants. The upgraded Analysis Settings keeps those ideas, you still guide how the AI interprets conversations, but expresses them as fields and rules you can reuse across assistants. Migration: Your existing Analysis Settings were migrated automatically into this new experience, the prior custom instructions carry forward as rules on the appropriate fields, so you should not need to recreate them by hand.
Destinations: Contact vs Conversation
Use destinations based on who or what should “hold” the value:
| Destination | Choose this when… |
|---|---|
| Contact | The value describes the person across visits (e.g., lead score, membership intent, tags). It belongs on their contact profile so it stays with them over time. |
| Conversation | The value is specific to this chat, text, email, or call (e.g., “reason for contacting us today”). Use when the answer might change the next time they reach out. |
System vs custom fields
| System fields | Custom fields | |
|---|---|---|
| Who manages them | Provided by the platform; core behaviors stay aligned with Replify. | You create and own them for your org’s workflow. |
| Editing field definition | You cannot edit the field definition—properties like name, key, type, where it saves, update mode, and timing are fixed for each system field. | You choose name, key, type, destination, update mode, and timing when you create the field and can edit them later (within product limits). |
| Rules | Each system field includes a default rule that defines baseline behavior. You cannot change that default itself; instead, you add rules that override it when you need instructions that better match your business or a subset of assistants—see Assigning rules. | You add rules and assign them org-wide or to specific assistants. |
If you’re unsure, start from system fields for standard signals and add custom fields only when you need something unique to your business.
Assigning rules: org default vs assistant override
For system fields, your override rules sit on top of the field’s built-in default rule. For custom fields, every rule is one you author. When you create or edit a rule, you choose how widely it applies:
- Org default — One canonical prompt for that field across your organization (unless an override applies).
- Assistant override — Applies only to specific assistants you pick. Use when one assistant needs different wording or criteria than the rest.
This is how you stop duplicating the same block of instructions on every assistant: write the rule once, assign it as the default, or to the assistants that need an exception.
Creating a new custom field (what each input means)
When you create a field, you’ll fill in details like the following.
| Input | What it’s for |
|---|---|
| Name | Display name used to find your Analysis Field more easily on the Analysis Settings (e.g., “Lead temperature”). |
| Key | A stable identifier for integrations and exports. This must be unique, and will be used in the Zapier integration along with the derived value. |
| Description | Optional context for your team so everyone knows what the field is supposed to capture. |
| Data type | What shape the value takes, e.g., text, number, yes/no, dates/times, or a structured object (think json). |
| Saves to | Contact or Conversation—see Destinations. |
| Update mode | Replace (overwrite), Add to existing (append when it makes sense), or Only if empty (Only write if not provided already). |
| Timing | Immediate vs Standard. Use Immediate when the value must be ready right away for automations (e.g., Zapier); Standard when a short delay is fine (generally less than 15 seconds after a conversation ends). You can only create up to 5 custom Immediate fields. |
After the field exists, you add rules (prompts) and assignments as described above.
How to use it (day-to-day)
- Open Settings → Analysis Settings, or go directly to Analysis Settings in the app.
- Use search and filters to find fields by source (System vs Custom), destination (Contact vs Conversation), or timing (Immediate vs Standard).
- Switch between Active and Inactive lists to manage fields you’ve turned off versus ones in use.
- Click a field to open its detail panel. For custom fields, edit field settings and rules as needed. For system fields, the field definition is read-only—use rules to layer overrides on top of the built-in default rule.
- Use Create field when you need a new custom field.
View rule output in the Inbox
The practical way to test a new or updated rule, especially a custom rule, is to run a new conversation the rule applies to, then open that thread in the Inbox. Analysis values appear under “Custom Analysis Fields”:
- In the Conversation Details section for that conversation, and
- In the Contact detail panel on the right (for values stored on the contact).
Use those spots to confirm the AI produced the output you expected before you rely on it in Zapier or downstream workflows.
Assistants tab
An Assistants tab may appear for future functionality that will continue to improve the experience.
Ideas for what to use it for
- Sales and routing — Capture qualification, urgency, or product interest once, reuse everywhere.
- Operations — Session-specific notes on the conversation vs durable traits on the contact.
- Automation — Push custom field values to Zapier for CRM updates, Slack alerts, or spreadsheet logging.
After big changes, run a few test conversations and check “Custom Analysis Fields” in the Inbox (conversation details and contact panel) so the right values land on the contact vs conversation, then verify any Zap you rely on.
What else you should know
Covering common questions so nothing important is buried:
- Turning fields off — You can disable custom fields when you no longer want them filled on future conversations; historical values will remain visible where they were already saved.
- Zapier — Custom fields are the usual hook for post-processing outside Replify, e.g. if something specific happens in a conversation, then handle that conversation differently, using these custom fields to route.
- Testing rules — There isn’t a separate “preview” required in Settings: add or edit your rule, complete a test conversation, then confirm values under “Custom Analysis Fields” in the Inbox—see View rule output in the Inbox.