Architecture
Agent Architecture
How Rox uses account context, reasoning, and connected tools to plan and execute work across the revenue lifecycle.
Rox is an outcome-oriented revenue agent that helps companies find, win, and grow business. It brings together context from across the business and external sources, reasons over that information, and takes action across the revenue lifecycle.
Rather than requiring users to coordinate separate tools and workflows, Rox provides a single agent experience that can understand a goal, determine the work required, and execute tasks using the relevant data and connected systems.
Context-aware reasoning
Every revenue task depends on context.
Preparing for a customer meeting, identifying expansion opportunities, and generating outbound pipeline require different information, but they all benefit from a consistent understanding of the account and its history.
Rox uses its System of Context to bring together relevant information from structured and unstructured sources, including CRM records, account activity, meeting transcripts, emails, documents, and external research.
The agent retrieves context relevant to the task, including information about the account, the seller, and the organization. This allows it to interpret requests in relation to existing relationships, previous interactions, and business objectives.
For more information, see System of Context.
Planning and execution
Rox translates natural-language requests into executable work.
When a user provides a goal, the agent interprets the request, identifies the information and capabilities required, and determines how to complete the task.
Execution generally involves three stages:
Interpret the request using the user's instructions and relevant business context.
Identify the research, analysis, content generation, or system actions needed to achieve the objective.
Use the appropriate tools and connected systems to complete the work and return results.
Complex requests may involve multiple steps and dependencies. For example, identifying expansion opportunities can require researching existing accounts, analyzing recent developments, evaluating potential fit, and preparing a recommended sales play.
Rox coordinates these activities within a unified agent experience, rather than requiring the user to manage each step independently.
Tools and connected systems
Rox uses tools to retrieve information, analyze data, generate content, and perform actions.
These capabilities span several categories:
Information retrieval: Access relevant account, contact, opportunity, and activity data from connected sources.
Research and analysis: Investigate companies, identify signals, evaluate opportunities, and synthesize findings.
Content generation: Produce account plans, research reports, emails, presentations, and other business artifacts.
Business actions: Perform supported operations through connected systems, including outreach and CRM updates.
Tools allow the agent to move beyond generating recommendations to carrying out the work required to act on them.
The available actions depend on the connected systems, configured capabilities, and applicable permissions.
Automations and ongoing execution
Not every revenue task begins with a new user request.
Rox Automations allow users to define recurring work that the agent can perform according to configured instructions and schedules.
For example, Rox can monitor accounts for relevant business developments, evaluate qualifying signals, and prepare research or recommended next steps.
Automations extend the agent's capabilities beyond individual conversations, allowing work to continue through configured workflows.
Users can create and manage Automations through natural-language interaction with Rox.
For setup instructions and examples, see the Automations documentation in the User Guide.
Human-agent interaction
Users interact with Rox through natural language, providing instructions, asking questions, and refining the work the agent produces.
Rox can return information and actionable outputs, including research summaries, generated documents, and workflow results.
Users can guide the agent by supplying additional context, changing requirements, or requesting further work.
For actions involving connected systems, execution is subject to the permissions and controls applicable to the user, integration, and workflow.
This allows Rox to support increasingly autonomous work while preserving the role of human judgment in revenue decisions.