Businesses are under pressure to move faster, operate more efficiently, and deliver seamless experiences across every function. Yet too much time is still spent on manual integration work, disconnected systems, and complex setup processes.
Generative Flows in MuleSoft for Flow: Integration
To address this, we’re introducing Generative Flows in MuleSoft for Flow: Integration, bringing generative AI and natural language directly into the integration experience. This makes it easier to automate processes, connect systems, and unlock value in minutes instead of weeks. By simplifying how integrations are built, we help organizations accelerate time-to-value, boost productivity, and scale automation across the enterprise.
Building automation workflows with MuleSoft for Flow is already simple and low-code, and now it’s even more powerful. With Generative Flows, users can describe what they need in natural language and instantly generate fully functional integrations. Instead of starting from scratch, teams can quickly create workflows that connect and orchestrate data across applications.
This removes setup friction, shortens implementation time, and enables teams to focus less on building integrations and more on driving business outcomes.
What is Agentforce for Flow?
Agentforce for Flowformerly known as Einstein for Flow, is Salesforce’s native generative AI tool that leverages large language models (LLMs) to power automation across the Salesforce ecosystem. The tool enables Salesforce admins, regardless of their programming experience, to easily create functional Salesforce Flows via natural language prompts. This removes the need for manual configuration, significantly streamlines and scales the automation development process, and helps users learn about Flow creation.
Faster impact with generative AI
Agentforce for Flow enables customers to achieve business outcomes faster. Instead of starting from a blank canvas or navigating complex configuration screens, citizen users can simply describe what they want to automate in natural language, and Agentforce for Flow does the heavy lifting.
- Accelerated time to value: Transition from a blank canvas to a functional and multi-system flow in real time. Build complex multi-system flows in seconds instead of hours or days.
- Empowered citizen developers: Salesforce Admins and business users can now orchestrate data across Zendesk, Jira, and QuickBooks conversationally using natural language.
- Reduced manual errors: AI interprets system relationships and suggests optimal mappings, reducing the risk of manual configuration errors.
- Smarter workflows: Beyond moving data, the engine can summarize and transform information in-flight, creating context-aware integrations.
What’s live and ready to use
Generative Flows allow users to move beyond manual mapping toward context-aware, adaptive automations. Whether syncing data within Salesforce or orchestrating workflows across multiple SaaS systems, Agentforce for Flow interprets natural language instructions and understands relationships between systems to generate desired integration logic. This release includes:
- Natural language to flow: Users describe the integration scenario in plain language. Agentforce AI selects the appropriate third party connectors, chooses relevant operations, and generates the required business logic
- Multi-connector support: Agentforce AI can orchestrate multiple systems within a single flow, such as Salesforce, Zendesk, and Outlook
- Connector static operations: Predefined connector operations with fixed inputs and metadata available at design time. The schema remains consistent and does not change based on configuration.
- Custom property editor and inline transform: Connector invocable actions can be configured using an intuitive custom property editor directly within the Flow Builder canvas
- Supported flow types: Generative Flows support a variety of Flow types, including Record-Triggered Flows, Autolaunched Flows, External System Change-Triggered flows, Screen Flows, giving users the flexibility to automate processes wherever they happen and unlock faster, more consistent outcomes across systems
Where Generative Flows drive business value
Use AI to draft flows for a wide range of business scenarios, from simple data synchronization to multi system orchestration. Explore common integration patterns below.
| Integration patterns | Prompt instructions | Systems involved |
|---|---|---|
| Outbound sync | When an opportunity is closed-won in Salesforce, create an Invoice in QuickBooks Online and mark it as ‘Unpaid’ | Salesforce and QuickBooks Online |
| Inbound sync | When a new customer is set up in Microsoft Business Central, create a corresponding account record in Salesforce | Microsoft Business Central & Salesforce |
| Cross system | Automatically update the Zendesk ticket to ‘Pending’ once a Jira issue is marked as ‘Done’ | Jira and Zendesk |
| Complex chain | When a Quickbooks Online Invoice over $5000 is paid, create a record in Google Sheets and notify me in Slack | QuickBooks Online, Google Sheets, Slack |
7 steps to help you draft flows in minutes
- Set up Agentforce Generative AI to get access to powerful generative AI features.
- Before you start prompting, ensure the Salesforce org environment is ready. You must pre-configure external system connections using third party connectors to allow Agentforce to access them. Configure these connections in Org → Automations → Integrations.
- Go to Automation Home Page > Click ‘New’ > Get Started with Automations > Let Agentforce Help You Build
- Describe instructions for the integration use case. A detailed prompt is better as it helps with the accuracy of the generated flow. See below on how to write the perfect prompt.
- Click Draft With AI
- Agentforce takes the context from the instructions and generates an integration flow in just a few seconds.
- After generating an integration flow with Agentforce, one can edit, save, test, and activate the flow just like any other flow in Flow Builder canvas. Like with anything created by AI, one should review the flow keeping a human in the loop to ensure accuracy prior to deployment.
How to write the perfect prompt
Prompts are the instructions that help LLMs generate content. To ensure high-quality output, a prompt should contain concrete information, context, and constraints. Vague or inaccurate instructions can lead LLMs to generate irrelevant or even incorrect responses.
While Agentforce for Flow handles the complex task of generating Flow metadata and connecting system APIs, accuracy of the result depends on the clarity of the instructions. High-quality output requires grounding to provide the specific context for business logic.
- Be specific with triggers: State exactly when the flow should start (eg “When a record is created in a system” or “When a specific field is updated in a source system object”)
- Define the chain of events: If you need multiple actions, use words like “and then” or “subsequently” to define the order
- Name system connections: Pre-configure external system connections in Salesforce Org → Automations → Integrations. Explicitly include the specific Connection Name in the prompt so Agentforce AI knows exactly which instance to use
- Mention field mappings: If you need specific fields synced (eg “Map Salesforce Opportunity Amount to QuickBooks Invoice Total”), include that in the instructions
Example of a quality prompt:
"When a new case is created in Salesforce, create a Zendesk ticket using the 'Create ticket' Zendesk operation. Once the ticket has been created in Zendesk, grab the Ticket id from Zendesk and update it in the triggering case record in the custom field Cust_Zendesk_Ticket_Id. Then, summarize the case content from the Salesforce case using subject and description and add it as a comment in Zendesk in the newly created ticket. The Zendesk connector must use the Zendesk1MSIC_Zendesk connection."The following image illustrates how the Agentforce Generative AI engine translates this natural language promptly into a structured, multi-step workflow.
The smart idea to integrate demo
Watch our demo to see how generative AI accelerates Flow creation using simple natural language prompts. In this walkthrough, we show how AI generated flows can be built in real time for two business scenarios.
The smart escalation with Salesforce and Jira
In this first use case, we show how the AI handles contextual logic. By simply describing the need to sync escalated cases to Jira, the AI:
- Determines the Flow Type: Automatically selects a Record-Triggered Flow.
- Sets the Criteria: Identifies the ‘Escalated’ status as the trigger filter.
- Builds the Logic: Automatically creates a concatenation formula (Case Number and Subject) to format the Jira comment.
Multi-path sync with Salesforce and QuickBooks Online
The second demo tackles a more complex challenge: the ‘Create or Update’ (Upsert) pattern. Watch as the AI:
- Designs the flow: Builds a Decision Element with two distinct paths (Create vs. Update)
- Maps complex data: Connects QuickBooks Customers to Salesforce Contacts, including custom field mappings that usually require careful manual configuration
- Ensures data integrity: AI chooses the correct ‘Update’ action based on the ID, ensuring no duplicate records are created
What’s next?
While this release delivers powerful generative capabilities, this is just the beginning. Our roadmap is focused on enabling richer conversational interactions, expanding connector coverage, and delivering a more guided connection setup experience. Available soon, even more exciting enhancements on the way include:
- Conversationally evolve a flow: Today, Agentforce AI generates the initial flow draft based on curated instructions. The upcoming Evolve a Flow experience will allow users to refine and modify flows conversationally using the Agentforce side panel.
- Connection management wizard: Planned improvements will provide a more seamless and guided experience for configuring external system connections, making connection setup more deterministic and user friendly.
- Connector dynamic operations: Support for operations that dynamically adapt based on the selected object or query.
- AI prebuilt flows: Pre configured, use case-specific natural language templates that allow users to quickly generate flows for common business scenarios. Users select a template and configure the required external connections through a guided setup experience.
With these capabilities, teams can move faster from idea to impact, turning natural language into real automation, reducing manual effort, and accelerating time-to-value across systems. By simplifying how integrations and workflows are built, we empower more users to innovate, streamline operations, and deliver measurable business outcomes at scale. Learn more about MuleSoft for Flow: Integration to boost business efficiency today.
