Salesboom is a cloud CRM platform whose Agent Management System (AMS) provides a governed environment for configuring, managing and supervising AI agents and reusable AI workflows that operate with approved business context and human oversight.
Salesboom AMS is the AI-agent governance layer within the Canadian Enterprise Stack. The system is intended to help organizations define what an agent is responsible for, what context it may use, which tools it may access, how its outputs are reviewed and how the workflow is monitored.
AMS is distinct from the Data Engine, which organizes governed business context, and the Integration Station, which provides scoped connectivity to supported business systems, APIs, webhooks, MCP tools and AI services.
Define agent purpose, responsibility, allowed business context, assigned workflows and the human owners responsible for supervision.
Govern reusable prompts, instructions, skills and workflow components so organizations can standardize how agent tasks are performed.
Control which CRM records, enterprise context, APIs or tools an agent may use according to implementation policies.
Configure review, approval or intervention points for sensitive actions, generated content and decisions where human control is required.
Maintain operational history for agent activity, workflow runs, approvals and outputs where included in the deployment.
AMS governance can include model-selection rules, usage limits and AI cost controls based on agent or workflow requirements.
Human-in-the-loop does not require every AI action to follow the same approval path. Instead, workflows can define where people must review, approve, correct, assign or stop agent actions. Low-risk research or drafting may follow one path, while customer-facing publication, financial actions or sensitive record changes can require additional controls.
This allows organizations to combine automation with accountable human decision-making instead of treating AI autonomy as an all-or-nothing choice.
The Salesboom Data Engine can organize approved CRM, ERP, accounting, website, partner and operational context. AMS can use permitted portions of that context to support research, drafting, analysis and workflow execution.
Example use cases from the Salesboom AI architecture include account research, sales-content drafting, questionnaire or proposal support, risk analysis, rate-card audits, demand forecasting, compliance checks, pipeline-risk forecasting, churn-risk detection and expansion-opportunity identification. Each use case should be scoped around the source data, model, workflow, permissions and required human review.
MCP means Model Context Protocol. It is an interoperability standard for connecting compatible AI systems to tools and context. It is not a proprietary protocol invented by Salesboom.
Salesboom can implement MCP servers, clients, wrappers or tools, alongside conventional APIs, web services and webhooks. The Hybrid API and MCP Integration layer explains the technical methods, while Integration Station owns the wider enterprise orchestration and connectivity architecture.
Configured language-model integrations can include supported OpenAI, Claude, Gemini and open-source model environments. The correct model depends on task requirements, security, cost, latency, context size, provider terms and deployment architecture.
AMS should therefore govern the model as one component of the workflow rather than assuming that every agent must use one vendor or one model for every task.
For the core residency and Canadian-service model, see Salesboom Canadian CRM with data hosted in Canada.
Salesboom CRM data, software, production servers and backups can remain in the Canadian Salesboom environment where the Canadian CRM deployment applies. That does not automatically extend Canadian residency to an external AI provider.
If prompts, files or selected business context are sent to OpenAI, Anthropic, Google or another model host, that processing follows the provider's infrastructure, regional configuration, retention policies and contractual terms. AI architecture should document what data may leave the Salesboom environment, which provider processes it, and which controls apply.
AMS is the control plane. The Agentic Workforce is the operating model in which people, AI agents, copilots and automated workflows work together across business processes.
This distinction keeps governance separate from business deployment: AMS manages and supervises agents; the Agentic Workforce describes how those governed agents participate in day-to-day work.
Salesboom AMS is a governed environment for configuring, managing and supervising AI agents and reusable AI workflows that use approved business context and configured tools.
Depending on implementation scope, AMS governance can include agent definitions, prompts, reusable skills and workflows, permissions, human approvals, logs, model connections and AI usage or cost controls.
Approved CRM and enterprise context can be organized through the Salesboom Data Engine and exposed to agent workflows according to permissions, workflow design and integration boundaries.
APIs, web services, webhooks and Model Context Protocol connections can be used through the Integration Station or other scoped integration methods to connect supported tools and systems to agent workflows.
Configured language-model integrations can include supported OpenAI, Claude, Gemini and open-source model environments. Availability, processing location, retention and contractual terms depend on the selected provider and deployment.
No. Salesboom CRM and approved CRM context can use Canadian-hosted Salesboom infrastructure, but prompts or context sent to an external model provider are processed according to that provider's infrastructure, region settings and contractual terms.