Agentic workflows, agents, and skills
Build Agent can generate agentic workflows, agents, and skills scoped to your custom app. Turn business requirements into configured AI artifacts without building from scratch.
You must install Now Assist for App Engine to create agentic workflows, agents, and skills. For more information, see Installing Now Assist for App Engine.
Skills must be published before they can be used as a tool by another skill.
- Skills:
- *FlowAction
- Now Assist Skill
- Script - Inline
- Script - Explicit
- *SubFlow
- WebSearch
- Agents:
- Catalog item: For more information, see Add a catalog item to an AI agent.
- Conversational topic: For more information, see Add a conversational topic to an AI agent.
- Flow action: For more information, see Add a flow action to an AI agent.
- Now Assist Skill: For more information, see Add a Now Assist skill to an AI agent.
- Record operations: For more information, see Add a record operation to an AI agent.
- Script: For more information, see Add a script to an AI agent.
- **Search retrieval: For more information, see Add a search retrieval to an AI agent.
- Subflow: For more information, see Add a subflow to an AI agent.
- WebSearch: For more information, see Add a web search to an AI agent.
- *You can use only FlowActions and Subflows that are already part of your app scope. Build Agent reuses existing FlowActions and Subflows instead of creating new ones.
- **When you create an agent that requires knowledge retrieval, Build Agent can configure a search retrieval tool scoped to your use case. The tool queries your existing search profiles to identify the best match, retrieves the associated search sources and return
fields, and checks whether semantic search indexes are available.Note:You must be on Zurich Patch 10 or higher to use the search retrieval tool.
The agentic capability turns business requirements into fully configured agents, skills, and agentic workflows for custom applications. It evaluates existing app context, such as tables, roles, business rules, and metadata, and applies a four-phase pattern to generate artifacts scoped to your application's data model, roles, and ACLs:
- Assess
- Evaluates data model readiness for AI.
- Analyze
- Identifies use cases from existing tables, roles, and flows.
- Design
- Proposes an architecture with a recommended quick-win.
- Build
- Generates agents and skills scoped to the application's data model, roles, and ACLs.
In-App agents can work with any custom scoped application. Applications with real data, established roles, and defined automation produce more targeted, domain-specific artifacts. Applications with no data or configuration produce agents that aren't scoped to a specific data model.