How to Build an AI Sales Copilot Using Salesforce, Gmail, and Slack Data
March 4, 2026
Sales teams generate valuable information across many platforms. Pipeline activity lives in CRM systems like Salesforce. Customer conversations happen in Gmail threads and Slack channels. Important context about deals and accounts is spread across all of them.
AI sales copilots aim to unify this data so that sales reps can ask questions, get summaries, and receive recommendations without switching tools.
The challenge is that building a copilot requires retrieving and combining data from multiple systems with different APIs, schemas, and authentication flows.
Unified simplifies this process by providing normalized APIs across CRM and messaging platforms. Instead of building separate integrations for Salesforce, Gmail, and Slack, developers can retrieve conversations, deal data, and account context through consistent objects and endpoints and use that data to power AI workflows.
Why SaaS Products Are Building AI Sales Copilots
AI sales copilots are becoming a core feature in many SaaS products. They help sales teams understand customer activity and take action faster.
Products building these capabilities include:
- Sales automation platforms
- Revenue intelligence tools
- CRM productivity extensions
- Customer success platforms
- AI sales assistants
These products rely on combining several types of data:
- CRM pipeline data such as deals, stages, and owners
- Email conversations between sales teams and prospects
- Internal messaging discussions in platforms like Slack
- Customer interaction history including meetings, calls, and notes
When these data sources are connected, an AI assistant can answer questions like:
- 'What's the latest activity on this deal?'
- 'Summarize the last email thread with this prospect.'
- 'Which deals are at risk this quarter?'
- 'What should I do next for this account?'
Common AI Sales Copilot Use Cases
AI copilots support a range of sales workflows.
Deal summaries
Automatically summarize opportunity history including emails, Slack conversations, and CRM activity.
Conversation insights
Analyze email and Slack messages to identify customer sentiment or risk signals.
Next-step recommendations
Suggest actions based on pipeline stage, recent activity, and communication patterns.
Customer conversation search
Allow sales teams to query historical conversations across Gmail and Slack.
Pipeline insights
Identify stalled deals or accounts that need follow-up.
Unified Categories Used for an AI Sales Copilot
An AI sales copilot typically retrieves data from three Unified API categories.
| Category | Description | Key Objects |
|---|---|---|
| CRM | Sales pipeline and account data from platforms like Salesforce | contact, company, deal, event |
| Messaging | Communication data from Gmail, Slack, and other messaging systems | message, channel |
| GenAI | Objects used to generate AI responses across LLM providers | genai_prompt, genai_model |
Using these normalized APIs allows developers to retrieve sales context from multiple systems without writing vendor-specific integrations.
Unified Objects and Key Fields
Unified standardizes the objects required for building AI assistants.
CRM Objects
| Object | Key Fields | Purpose |
|---|---|---|
| Contact | name, emails, company_ids | Customer profile information |
| Company | name, domains, industry | Account-level context |
| Deal | amount, stage_id, probability, closing_at | Pipeline and opportunity data |
| Event | type, created_at, contact_ids | Sales activity such as calls or meetings |
Messaging Objects
| Object | Key Fields | Purpose |
|---|---|---|
| Message | message, subject, created_at, author_member | Email or chat content |
| Channel | id, name, members | Context for Slack or messaging conversations |
| Thread | parent_id, message_thread_identifier | Conversation history across replies |
Messaging records also contain:
- sender information (
author_member) - recipients (
destination_members,mentioned_members) - message bodies (
message,message_html) - attachments and links
- timestamps for conversation timelines
These fields provide the context that AI models use to understand customer conversations.
GenAI Objects
Unified's GenAI API normalizes LLM interactions across providers.
| Object | Purpose |
|---|---|
| genai_prompt | Sends prompts to a language model and returns responses |
| genai_model | Describes available AI models and their capabilities |
These objects allow applications to generate AI responses without changing code for each provider.
Connecting Customer Integrations
Customers connect their CRM and messaging platforms using Unified Connect.
Typical process:
- Your application launches the Unified Connect authorization component.
- The user selects integrations such as Salesforce, Gmail, or Slack.
- The user authorizes access to their account.
- Unified redirects back to your application with a connection_id.
Your application stores the mapping:
user_id → connection_id
All Unified API requests use this identifier to retrieve data from the authorized integration.
Retrieving CRM and Messaging Data
AI copilots retrieve context from CRM and messaging endpoints.
CRM endpoints
GET /crm/{connection_id}/contact
GET /crm/{connection_id}/company
GET /crm/{connection_id}/deal
GET /crm/{connection_id}/event
These endpoints provide pipeline information, account data, and sales activity.
Messaging endpoints
GET /messaging/{connection_id}/message
GET /messaging/{connection_id}/channel
Messages include the full conversation body, sender information, recipients, timestamps, and thread relationships.
Threads can be retrieved by filtering messages using the parent_id parameter and expanding descendants.
This allows an AI assistant to retrieve full email or Slack conversation history.
Building a RAG Pipeline for Sales Context
Most AI copilots rely on retrieval-augmented generation (RAG) to provide accurate answers.
A typical architecture follows these steps.
1. Connect integrations
Customers authorize CRM and messaging systems through Unified Connect.
2. Subscribe to webhooks
Register webhook subscriptions for objects such as messages, deals, and contacts.
3. Retrieve updated records
When a webhook fires, retrieve the latest object from the Unified API.
4. Chunk conversation content
Split long email threads or Slack conversations into smaller segments.
5. Generate embeddings
Use an embedding model to convert text into vector representations.
6. Store vectors in a database
Persist embeddings in a vector database such as Pinecone or pgvector.
7. Retrieve relevant context
When a user asks a question, search the vector database for relevant records.
8. Generate an AI response
Send retrieved context to the language model to generate a grounded answer.
This pipeline allows AI assistants to answer questions using real customer data.
Generating AI Responses
Unified's GenAI API standardizes how applications interact with language models.
To generate a response, create a genai_prompt object containing:
- a list of messages
- optional model parameters
- generation controls such as temperature or max tokens
Example request structure:
{
"messages": [
{ "role": "system", "content": "You are a sales assistant." },
{ "role": "user", "content": "Summarize recent activity for this deal." }
]
}
The API returns a prompt object containing:
- generated responses
- token usage statistics
Applications can also retrieve available models using the genai_model endpoint, which provides model IDs, token limits, and provider information.
Unified supports multiple LLM providers, allowing developers to switch models without changing application logic.
Keeping the Copilot Updated
AI assistants require fresh data to remain useful.
Unified provides webhook events that notify applications when records change.
Typical webhook events include:
CRM events
- contact created / updated / deleted
- deal created / updated / deleted
- pipeline updates
Messaging events
- message created
- message updated
- channel created or deleted
- interaction events such as reactions or mentions
Unified supports both native webhooks and virtual webhooks.
Native webhooks deliver immediate push notifications when supported by the provider. Virtual webhooks simulate real-time updates through managed polling when native events are unavailable.
These events ensure the AI copilot always has access to the latest CRM activity and customer conversations.
Supported Platforms
Unified provides broad integration coverage across CRM and messaging systems.
CRM platforms (47+ integrations)
Examples include:
- Salesforce
- HubSpot
- Pipedrive
- Zoho CRM
- Microsoft Dynamics
- Copper
- Close
Messaging platforms
Examples include:
- Gmail
- Slack
- Microsoft Teams
These integrations allow AI assistants to access sales conversations and pipeline data across many customer environments.
Why Product Teams Choose Unified
Building an AI sales copilot requires access to multiple sources of customer context.
Without integration infrastructure, developers must maintain separate integrations for each CRM and messaging platform.
Unified simplifies this by providing:
- normalized APIs across CRM and messaging platforms
- real-time data access and webhook events
- consistent authentication flows through Unified Connect
- AI-ready data models for building RAG pipelines
Instead of managing integration complexity, product teams can focus on building AI capabilities that help sales teams close deals faster.
Start building AI sales assistants powered by Salesforce, Gmail, Slack, and dozens of other integrations today.