Purpose Built agents

Purpose Built agents are system agents Loyal Guru publishes. They exist so you can draft a discount or a points campaign, look up Company data, or ask how the product works — without building an agent first.

Aligned with AI Loyalty Operations: campaign prompting turns ideas into execution, and you interact with the loyalty program through natural language.

You do not create, rename, or instruct them. Pick an Active one in the Owner header chat. The Company only sees the system agents Loyal Guru has published for it.

For custom agents, see Create your own agents. Overview: Agentic layer.

What they are for

Three jobs, each with a specialist:

Job	Value	System agents
Draft a campaign	Turn a brief (or a flyer) into a valid campaign payload you can review and save	Promotions Agent, Points Agent
Look up data	Answer from the Company’s own locations, catalog, customers, scores, coupons, and vouchers	Loyal Guru API Agent, Customer Agent
Explain the product	Fields, validations, and how loyalty and promotions work — from public docs	Knowledge Base Agent, Guru Agent

You stay in the loop. A purpose-built agent does not launch Targeted Delivery or send personalized coupons. That product is AI 1-to-1 Delivery.

Published system agents

These are the system agents Loyal Guru publishes today. Use the one that matches the question.

Promotions Agent

Drafts a discount campaign from natural language. It can also read a promotional image (flyer, poster, screenshot) when the chat accepts an image. It builds the campaign JSON and validates it before showing it to you.

Ask it when you want to create a discount, for example:

• 10% off fruit in store X this weekend
• NxM or strikethrough from this flyer
• Check that this discount payload is valid before I save it

Points Agent

Drafts a points campaign (current engine) from natural language. It builds the campaign JSON and validates it before showing it to you.

Ask it when you want to create a points campaign, for example:

• 1 point per €1 on the grocery category
• Double points on Tuesdays in these stores
• Check that this points campaign payload is valid before I save it

Loyal Guru API Agent

Reads Company data exposed by Loyal Guru: locations, customers, products, product categories (features), coupons, vouchers, and related entities.

Ask it when you need a catalog or setup fact, for example:

• Which locations are in Ireland?
• List product categories I can target
• Which coupons or vouchers exist for this campaign?

Customer Agent

Reads customer data: search or list customers, profiles, loyalty scores, score transactions, vouchers, and recent purchases.

Ask it when the question is about a member, for example:

• Find this customer and show their available points
• What did they buy recently?
• Which vouchers do they have?

Knowledge Base Agent

Answers how Loyal Guru works: campaign fields, validation rules, product concepts, and general questions. It uses the public documentation (search and Q&A over indexed docs). It also handles small talk.

Ask it when you need the product explained, for example:

• What does stackability mean on a discount?
• Which fields are required to save a points campaign?
• How is a voucher different from a coupon?

Guru Agent

Generalist. Use it when the question mixes topics, or you do not know which specialist to open. For a focused draft (discount vs points) or a data lookup, prefer the agent above.

What you can configure

You do not create system agents in the form. You choose which conversation is active:

Lever	What you decide
Default chat	Stay on the built-in conversation (no system or custom agent selected).
System agent	Pick a Loyal Guru agent from the list or the header dropdown.
Search	Filter the list by name, description, or creator.

How it works

Default chat

When no system or custom agent is selected, the panel uses default chat. Use it for a quick question, or switch to a purpose-built agent when you know the job (draft a discount, draft points, look up a customer).

A selected custom agent replaces default starters with its own conversation starters when you saved any.

Purpose Built vs Create your own in the list

Open Agents from the panel. Each row shows name, description, creator, and status.

Signal	Purpose Built	Create your own
Creator	Created by the system (no user name)	An Owner user with a name
Who can create it	Loyal Guru	An Owner user with Create on the Agents entity
Chat	Only when status is Active	Only when status is Active

The header dropdown lists up to five Active agents so you can switch without opening the full list. Use Default to go back to the built-in chat.

Status

An agent is not chat-ready until Loyal Guru finishes publishing it.

Status	What you can do
Pending / Processing / Submitted	Wait. Chat is closed.
Active	Start chat.
Failed	Chat is closed. Retry from edit if you have permission, or contact Loyal Guru.
Deleting	Chat is closed. The row disappears when delete finishes.
Delete failed	Chat is closed. Try delete again or contact Loyal Guru.

Examples

Simple. Open Promotions Agent and ask: “Draft 10% off the fruit category in Dublin stores next weekend.” Review the payload, then save the discount in Owner.

Combined. Open Points Agent for “2 points per €1 on coffee, Tuesdays only.” If you need store codes first, ask Loyal Guru API Agent, then go back to Points Agent with those locations.

Edge. The agent is still Processing. The panel will not send messages until status is Active. If the Company does not have discounts or points contracted, say so and contact your CSM — do not invent a campaign type you cannot save.

Configuration

1. Confirm the Company has AI agents available.
2. Confirm the Owner user can open AI chat and list agents.
3. Open the AI chat control in the Owner header.
4. Open the agent list and select a system agent whose status is Active (Promotions, Points, API, Customer, Knowledge Base, or Guru).

System agents are not edited or deleted from the custom-agent form. For a Company-specific helper, create your own agent.