18 Launch AI Agents

Offer Similarity Detection AI Agent

Product proliferation, especially for bundles, is one of the most significant challenges faced by Communications Service Providers (CSPs). The inability to determine whether a similar bundle already exists before launching a new one often results in the creation of multiple offers with little or no differentiation. This can lead to a poor customer experience, increased operational overhead, and greater complexity in offer management. The ability to identify similar offers before introducing new ones to the market promotes reusability, simplifies global changes, and helps mitigate product proliferation.

You can detect similar offers (bundle and atomic offers) while creating, revising, or cloning offers by using the Fusion AI Agent. Similarity is determined based on multiple aspects of an offer, including the bundle structure, mandatory and optional components, option groups, attributes, pricing strategies, selling rules, and terms and conditions, including commitment terms with or without penalties.

Similar offers of the same offer type are presented and ranked according to their semantic similarity as Highly Similar, Mostly Similar, or Partially Similar. You can download the results to view a detailed, business-friendly explanation of the similarities and differences identified for each matching offer.

The similarity scoring and reasoning include:
  • Bundle/package pricing, rules, and terms
  • Package commitment terms
  • Bundle component-level pricing
The View Similar Offers action, which invokes the AI Agent, is available from the offer details page when you open it from the offer dashboard. You can also download a detailed explanation of the similarities and differences identified among the bundle offers.
Offer Similarity Detection involves three agents:
  • Similarity Ingestion Assistant: This seeded agent prepares offers for similarity searches. It retrieves eligible offers and generates structured summaries using predefined prompts. The agent then writes these summaries to vector storage by using the Vector Writer tool.
  • Similarity Detection Assistant: This seeded agent identifies and ranks similar offers. It reads the summarized input and queries vector storage by using the Vector Reader tool to retrieve and rank similar offers.
  • Similarity Cleanup Assistant: This template agent maintains index consistency. It retrieves all related vectors and removes stale entries (for example, offers that have been deleted through the user interface). It then resets ingestion as needed to ensure that the index remains accurate and up to date.

Offer Design Integrity Check AI Agent

One of the major challenges faced by Communications Service Providers (CSPs) is incorrect order capture, which can lead to order fallout, provisioning failures, and incorrectly billed services. In many cases, bundle offers contain entitlements that violate selling rules, have incorrect validity periods, or do not fully satisfy the intended business requirements. These issues increase support costs because corrections must be made and propagated across multiple systems. They can also result in indirect business impacts, such as reduced customer satisfaction, lower Net Promoter Scores (NPS), and increased customer churn.

The Offer Design Integrity Check AI Agent enables CSPs to design and release offers with greater confidence by proactively identifying potential design inconsistencies and policy violations. This helps reduce operational costs, accelerate time to market, improve customer experience, and increase operational reliability.

Offer Design Integrity Check AI Agent performs the following validations:
  • Validates incompatible component products or services within a bundle offer.
  • Validates the effective dates of all component products against the effective dates of the bundle offer.
  • Validates that the intent described in the offer description aligns with the components included in the bundle offer.
  • Validates the starting price of the bundle offer.

You receive both a summary and a detailed explanation of the findings, including actionable recommendations and suggestions. In addition, the AI Agent presents similar offers based on the structure and attributes of the bundle offer.

The Check Integrity action, which invokes the AI Agent, is available from the offer details page. You can also click the Check Integrity control when creating or updating an offer to invoke the Design Integrity Check AI Agent.

When invoked, the AI Agent validates the offer and classifies it as Valid or Invalid. It also provides a detailed explanation of the validation results, including the factors that make the bundle invalid, potential issues, warnings, recommendations, and suggested corrective actions.

Results of the Design Integrity Check AI Agent

In addition to providing recommendations and suggestions to help make a bundle offer valid, the AI Agent also identifies semantically similar bundle offers, enabling Product Managers to make informed decisions and take appropriate actions.

All findings are grouped into the following easy-to-understand categories:
  • Critical: Issues that must be resolved before the rule can be saved or confirmed.

    Note:

    This validation can be disabled.
  • Warning: Issues that have the potential to cause failures during order capture, billing, or provisioning.
  • Recommendation: Suggestions for improving the offer design and overall quality of the bundle offer.

In addition, you can validate rule creation for any overlapping or contradictory eligibility or compatibility conditions for the same subject and parameter combination that would cause a clear conflict.

You can also configure additional validations for the Offer Design Integrity Agent by copying the Offer Integrity Advisor template available in Oracle AI Agent Studio. This enables organizations to extend the out-of-the-box offer integrity validations with customer-specific governance rules and validation logic tailored to their business requirements.

The Offer Design Integrity Agent evaluates offer payloads, bundled components, eligibility rules, compatibility rules, recommendation rules, lifecycle dates, and other contextual information before offer publication. Organizations can configure additional AI-driven validations to identify invalid configurations, enforce catalog governance policies, and improve offer quality.

In addition to the out-of-the-box validations delivered through the two nodes, the Offer Integrity Advisor Workflow node, which runs the default validations, and the Summarizer node, which curates the output from the validation node for presentation in the user interface, you can copy the agent template to create your own agent.

Using Fusion AI Agent Studio, you can:
  • Create custom validation nodes within the Offer Integrity Advisor workflow.
  • Extend seeded offer integrity validations with organization-specific business rules.
  • Configure AI-driven validation prompts using contextual offer payload data.
  • Classify validation results as Critical, Warning, or Recommendation.
  • Generate structured validation outputs by using Specification Mode.
  • Configure summarization logic to consolidate seeded and custom validation results.

Rules Integrity Check AI Agent

Communications Service Providers (CSPs) manage what is sold, to whom, and how through a complex set of selling rules.

These rules broadly fall into the following categories:
  • Eligibility rules such as channel, location, customer segment, and other qualification criteria.
  • Validation of compatible offerings within bundles.
  • Cross-sell and upsell recommendations presented during the ordering process.
  • Plan upgrade and downgrade rules.
  • Solution-level rules such as constraints rules applied to B2B quotes.

CSPs face significant operational and financial risks when catalog and selling rules across eligibility, compatibility, recommendations, migrations, and solution constraints become contradictory or inconsistent. Such conflicts can result in incorrect orders, fulfillment failures, billing inaccuracies, poor customer experiences, and increased costs associated with order fallout and jeopardy management.

Additionally, managing evolving rules across channels, bundles, customer segments, and lifecycle states becomes highly error-prone without intelligent semantic validation, proactive risk detection, and AI-driven recommendations to prevent conflicts before offers and rules are launched.

Rules Integrity Check AI Agent performs the following validations:
  • Detects rule conflicts and inconsistencies within and across rule types, including:
    • Compatibility
    • Constraint
    • Eligibility
    • Recommendation
  • Ensures valid product bundling by identifying:
    • Incompatible combinations
    • Contradictions between compatibility and constraint rules
  • Validates the alignment of recommendation rules with:
    • Compatibility
    • Constraint
    • Eligibility
  • Identifies contradictions:
    • Within the same rule type
    • Across different rule types
  • Use semantic and probabilistic validation techniques, rather than relying solely on deterministic checks
  • Recommends corrective actions and safer alternatives
  • Supports sequence-aware validation by considering the order in which rules are created and modified. For example:
    • An existing incompatibility rule blocks or flags a recommendation rule
    • An existing recommendation rule flags new exclusion rule
  • Provides actionable outputs, including:
    • Conflict identification
    • Detailed explanations of detected issues
    • Suggested resolution paths

You receive both a summary and a detailed explanation of the findings, including actionable suggestions and recommendations.

The Check Integrity action for invoking the AI Agent is available from the Rules page. You can also click the Check Integrity control while creating or updating an offer.

When invoked, the AI Agent validates the rule and marks it as valid or invalid. In addition, it provides an explanation of why the rule is invalid, along with details of potential issues, warnings, and recommendations for addressing them.