Templates

A template is a pre-built AI Employee that you clone into your own Workspace and customize. Instead of assembling a workflow agent by agent, you start from a known-good blueprint, connect your data, adjust a few Instructions and categories, and publish. Ema ships a set of curated templates (currently spanning support, sales, HR, and IT & Ops) and you can also save any of your own AI Employees as a template to reuse across your team.

Library presets and existing workflows. Search & Respond, Knowledge Search, Search Query Planner, and Respond are no longer offered in the Add an Agent library. Existing workflows and templates that use these types continue to run. For new search-and-answer steps, add a Custom Agent with a Knowledge Base; for response-only steps, add a Custom Agent without a Knowledge Base.

What a template is

Every template carries the same building blocks as a hand-built AI Employee:

  • A workflow — a directed graph of typed agents (a start node, one or more agent nodes, and a publish node) wired together with edges. The curated templates use the configurable agent types intent_classification, search_respond, extraction, rule_validation, and custom, plus respond for free-form chat replies and document_agent for longer document generation. See the Agent reference for every agent type.
  • Starting Instructions — each agent node ships with a system prompt and, where the agent type needs it, a type_config (for example, the intent categories for an intent_classification agent, or the output schema for an extraction agent).
  • A starting configuration — model selection (EmaFusion™ "Let Fusion Pick" by default), an optimization priority, feedback settings, and, for chat templates, a welcome message and conversation starters.
  • Recommended Knowledge Bases — guidance on which Knowledge Bases to connect for the best results.

A template is not connected to your live data when you clone it. It is a launch-ready starting point: you connect your Knowledge Bases and Tools, adjust the Instructions, and publish.

Clone, not link. Creating an AI Employee from a template copies the template's workflow into a new, independent draft. There is no ongoing link. Later changes to the template do not flow back into AI Employees you already created, and your edits never change the template.

Curated templates

Ema ships a set of curated templates, shown with the Ema source label in the picker:

TemplateCategoryTypeWhat it does
Customer Support ChatSupportChatClassifies the customer's issue, searches your Knowledge Base, and generates a cited response across multi-turn conversations.
Agent AssistSupportChatSuggests real-time resolutions to human agents by classifying the query and searching internal knowledge.
Sales Development RepSalesChatGenerates personalized outreach grounded in product docs and case studies.
Employee OnboardingHRChatAnswers new-hire questions by classifying the topic, searching HR docs, and composing an answer.
Policy AssistantHRChatA focused Knowledge Base search across HR policies, returning cited policy sections.
IT HelpdeskIT & OpsChatClassifies IT tickets, searches runbooks, and resolves or escalates.

Each template has a reference page covering what it does, the workflow and agents it ships with, its starting configuration, and how to create an AI Employee from it. See Template reference pages below.

Naming. In the product, curated templates use short functional names (for example, "Customer Support Chat"). Throughout this documentation we refer to the AI Employee you create from a template using the "[Name] AI Employee" pattern — for example, the Customer Support AI Employee is the AI Employee you build from the Customer Support Chat template.

Creating an AI Employee from a template

  1. Go to the AI Employees list (the /ai-employees route).
  2. Select Build an AI Employee. The template picker opens.
  3. Browse or search the templates. Use the category chips — All, Support, Sales, Marketing, HR, IT & Ops, Finance, Other — to filter, or type in the search box.
  4. Select a template card to open its detail view. The Overview tab shows a workflow preview, the value proposition, and how-to-use steps. The Configuration tab (when the template includes a configuration) shows the model selection, optimization priority, feedback settings, welcome message, and recommended Knowledge Bases.
  5. Select Use template. Ema clones the template into a new draft AI Employee and opens it in the AI Employee builder, where you finish configuring it.

To build a custom AI Employee instead of starting from a template, select the Start from scratch card in the picker. This opens the create dialog and gives you a blank workflow.

After cloning, work through the template's How to use steps in order. They tell you which Knowledge Bases to connect, which categories or schema fields to adjust to match your taxonomy, and which Instructions to align to your brand voice before publishing.

What happens on the API

The template picker calls the workflow service. Creating an AI Employee from a template is a single request:

POST /api/v1/workflow/templates/{id}/create-workflow
Content-Type: application/json

{ "name": "My customer support AIE" }

The response is a new draft Workflow with the template's DAG definition, interaction type, and metadata copied into it. The new workflow is independent of the template. To list the templates available to you, the picker calls GET /api/v1/workflow/templates; starter (blank-modality) templates are hidden from this list and served separately via GET /api/v1/workflow/templates/starters.

Saving your own template

Any AI Employee you build can become a reusable template for your team. From the AI Employee, open Save as template and provide:

  • A name and description.
  • A category (Support, Sales, Marketing, HR, IT & Ops, Finance, or Other).
  • How to use — up to four numbered steps that guide whoever clones the template.
  • A value proposition — up to four benefit points.

Saving creates a custom template (shown with the Custom source label) scoped to your Workspace. Under the hood this is:

POST /api/v1/workflow/templates
Content-Type: application/json

{
  "workflow_id": "…",
  "name": "Returns specialist",
  "category": "Support",
  "how_to_use": "1. Connect your returns policy KB. 2. …",
  "value_prop": "Cuts returns handling time. …"
}

Custom templates can be edited (PUT /templates/{id}) and archived (DELETE /templates/{id}) by their creator, a system_admin of the owning Workspace, or an env_admin. Curated (Ema-shipped) templates cannot be modified or archived.

Templates with sub-AI Employees

When the AI Employee you are saving references other AI Employees via call_aie nodes, the template stores a full export bundle rather than a flat DAG snapshot. A flat snapshot keeps the original target_workflow_id values, which point to AI Employees in your Workspace that do not exist in the destination Workspace, so spawning from that snapshot would produce a broken AI Employee with missing sub-AI Employees. The bundle embeds each sub-AI Employee as a nested bundle; when someone creates an AI Employee from the template, the instantiation step creates those sub-AI Employees in the new Workspace automatically.

Voice AI Employees are not affected by this: their configuration lives in voice-v1 and is stored on the config snapshot, not in a bundle.

Sharing templates across Workspaces

If your organization has a Workspace hierarchy, a system_admin of a parent Workspace can grant a custom template to a descendant (child) Workspace. The child Workspace's builders then see the granted template in their picker alongside their own custom templates and the curated set. Grants are managed through the template grants endpoints (GET/POST /templates/{id}/grants and DELETE /templates/{id}/grants/{tenantId}). Only the owner Workspace can grant or revoke.

Template reference pages

Every curated template has a reference page, grouped below by category.

Support

Sales

  • AI SDR — generate personalized, knowledge-grounded sales outreach.

HR

IT & Ops

Last updated: Oct 12, 2026