Real-time Personalized Sales Guidance
This challenge involves developing a sophisticated multi-agent system using the Claude Agents SDK to provide personalized, real-time sales guidance. The system will act as an intelligent sales assistant, offering deal-specific recommendations, objection handling strategies, and next-step actions to sales representatives. The core of this system will be its ability to understand complex sales conversations and leverage extensive knowledge for strategic advice. Developers will design agents with extended thinking capabilities using Claude 4 Sonnet, which will then interact with a specialized 'Knowledge Agent' powered by Llama 4 Maverick for in-depth data retrieval and synthesis from various sales enablement resources. The system will integrate with a secure access platform like Aembit to ensure controlled and auditable access to sensitive CRM data. Libretto will be used for intelligent model routing and A/B testing of different guidance strategies, optimizing performance. Bito AI will serve as the conversational interface for sales reps, providing instant, context-aware advice. The overall agent orchestration will leverage Letta AI's capabilities for managing agent lifecycles and tool orchestration.
What you are building
The core problem, expected build, and operating context for this challenge.
Build a Claude Agents SDK multi-agent system for personalized sales guidance, utilizing Llama 4 Maverick for knowledge synthesis, secure data access, and a conversational interface for sales reps.
How work is evaluated
The evaluation will assess the system's ability to generate relevant and personalized sales guidance, handle common sales objections, securely access data, and provide an effective conversational experience.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
GuidanceRelevance
The generated guidance is contextually relevant and actionable (80% relevance score).
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
ObjectionHandlingEffectiveness
The objection handling scripts are coherent and strategically sound.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
SecureAccessCompliance
Aembit integration successfully enforces access policies.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
BitoAIResponseQuality
Bito AI provides accurate and helpful responses to sales queries.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
SalesScenarioCoverage
Percentage of test sales scenarios for which meaningful guidance was provided. • target: 0.85 • range: 0.7-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
ModelRoutingEfficiency
Latency reduction through Libretto's model routing compared to direct calls (ms). • target: 50 • range: 0-100
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Master the Claude Agents SDK for constructing hierarchical multi-agent systems capable of complex decision-making and tool use in dynamic environments.
Develop agents with advanced extended thinking logic using Claude 4 Sonnet to interpret sales contexts, anticipate objections, and formulate strategic responses.
Implement a specialized 'Knowledge Agent' powered by Llama 4 Maverick, focused on synthesizing sales collateral, customer histories, and market data for actionable insights.
Design and integrate secure data access tools using Aembit to enforce least-privilege policies for CRM and other sensitive enterprise systems.
Orchestrate intelligent model routing and A/B testing with Libretto to dynamically select the best LLM (e.g., Claude 4 Sonnet or Llama 4 Maverick) for specific sales tasks.
Build and deploy a real-time conversational sales assistant using Bito AI, providing an intuitive interface for sales professionals to receive guidance and insights.
Reference links and supporting material
Sample data for 3 tasks
Challenge setup
The evaluation will assess the system's ability to generate relevant and personalized sales guidance, handle common sales objections, securely access data, and provide an effective conversational experience.
Challenge input
JSON object containing 'deal_context' (text), 'conversation_history' (array of strings), and 'objection_type' (string).
Guidance
Framework
Zed
High-performance code editor
Claude 4 Sonnet
Policy Serving in the agent workflow.
Evaluated output
JSON object with 'recommended_action' (string), 'objection_handling_script' (string), and 'relevant_resources' (array of strings).
- The generated guidance is contextually relevant and actionable (80% relevance score).
- The objection handling scripts are coherent and strategically sound.
- Aembit integration successfully enforces access policies.
- SalesScenarioCoverage target: 0.85
- 3 public reference cases
- Python execution harness
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Guidance · Required
- Zed · Optional
- Claude 4 Sonnet · Optional
Evaluation contract
- GuidanceRelevance · Weight 1
- ObjectionHandlingEffectiveness · Weight 1
- SecureAccessCompliance · Weight 1
- BitoAIResponseQuality · Weight 1
- SalesScenarioCoverage · Weight 1
- ModelRoutingEfficiency · Weight 1
Metrics
- SalesScenarioCoverage · Target 0.85
- ModelRoutingEfficiency · Target 50
Reference cases
- 3 public reference cases
Workflow tasks
- Task 1
Input: JSON object containing 'deal_context' (text), 'conversation_history' (array of strings), and 'objection_type' (string).
Output: JSON object with 'recommended_action' (string), 'objection_handling_script' (string), and 'relevant_resources' (array of strings).
Execution harness: Python
- Task 2
Input: JSON object representing a request for specific customer data from the CRM, along with simulated agent identity and permissions.
Output: JSON object containing 'data_retrieved' (boolean), 'error_message' (string, if any), and 'access_policy_checked' (boolean).
Execution harness: Python
- Task 3
Input: Array of strings, each representing a user query to Bito AI.
Output: Array of strings, corresponding responses from the Bito AI interface.
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.