Using EKWIS AI inside Salesforce Flow

EKWIS AI is a managed Salesforce package that gives admins and developers reusable Flow actions for controlled AI work. It is available by invitation to existing EKWIS customers as a software-as-a-service subscription.

The product removes a repeated technical job: building and maintaining the same callout, request and response handling whenever a Salesforce process needs an AI-generated value. The customer still needs a defined use case, approved provider configuration, testing and clear ownership of what Flow does with the result.

Three output patterns

Different business processes need different levels of control. EKWIS AI separates field-oriented updates from general response handling rather than treating every request as the same action.

• Record Updater actions prepare a value for a target Salesforce field and return it to the surrounding Flow.
• Responses Service actions call the configured provider through Salesforce credential infrastructure and return content that Flow can route or pass to another step.
• Structured responses return predictable fields for use cases where Flow needs to validate and act on defined output rather than free text.

Start with a contained operational use case

Start with a use case that has a clear input, an output that can be checked and a limited consequence if the answer is wrong. Examples include summarising a case for an internal handover, classifying an inbound request, preparing a draft field value or extracting defined values from text for later review.

• Define the record, fields and related data the action is allowed to receive.
• Specify the expected output and what counts as an unacceptable result.
• Decide whether a person must review the response before a record changes.
• Handle timeouts, provider errors, empty responses and invalid structured output explicitly.
• Measure output quality on representative records before widening the automation.

Why Flow is the operating layer

Flow already contains much of the context around Salesforce work: record-trigger conditions, decisions, approvals, assignments, notifications and updates. Exposing the AI capability as an action lets teams keep those controls in a familiar place.

• Admins can use established Flow patterns for contained automation.
• Developers can review or extend the design when the use case needs Apex, additional security controls or more complex orchestration.
• Validation and approval steps can sit between the response and the final record update.
• The same packaged action can support different processes without copying the integration code into each one.

Controls to design around the package

Installing a managed package does not decide what data should leave Salesforce or when AI output is safe to use. Those decisions belong to the implementation.

• Provider and credential configuration, including who can view or change it.
• Field-level access and the minimum record context sent with each request.
• Prompt ownership, version control and regression testing.
• Logging that supports diagnosis without exposing sensitive prompt or response data.
• Human review for material customer, financial, security or operational decisions.
• A fallback path when the provider is unavailable or the response does not pass validation.

Availability and guided adoption

EKWIS AI is available by invitation to existing EKWIS customers rather than as a self-service product. The guided release combines the package with help defining the use case, configuring the Salesforce automation and reviewing how it behaves in the real process.

EKWIS AI assists with defined work inside controlled Salesforce processes. People remain accountable for process design, security, testing and release decisions.