This article summarizes and analyzes a long-form live session hosted by 9x on building AI agent workforces using Relevance AI. The session covered the conceptual shift from copilots and deterministic automations to autonomous, role-based AI agents, and demonstrated how multiple agents can collaborate in production workflows.
TL;DR
- Problem: Traditional automations and AI copilots still require constant human prompting and do not scale domain expertise well.
- Solution: Relevance AI enables autonomous agents with instructions, tools, and knowledge that can collaborate as a workforce.
- Outcome: Teams can operationalize research, outreach, content, and operations with modular, maintainable AI systems.
Context of the session
The session was part of a 9x live workshop series focused on AI and automation skills for business professionals. The guest speaker from Relevance AI presented both conceptual foundations and live demonstrations of building single agents and multi-agent systems.
The discussion positioned AI agents as a third layer beyond software and copilots. Software scales poorly without headcount. Copilots improve individual productivity but still rely on human direction. Agents are designed to operate autonomously within defined roles.
What defines an AI agent in Relevance AI
An AI agent in Relevance AI is built around three core components.
- Instructions: The role, objective, decision logic, and boundaries of the agent, written in plain language.
- Tools: Integrated capabilities such as web search, CRM actions, email, APIs, and internal utilities.
- Knowledge: Contextual data sources such as documents, tables, or reference material used during reasoning.
This combination allows agents to act dynamically rather than follow fixed step-by-step flows. Agents decide which tools to use and when, based on their objective.
Agents versus deterministic automation
A key theme of the session was the contrast between agentic systems and traditional automation platforms.
Deterministic tools require predefined chains and break when assumptions fail. Agents, by contrast, can adapt, retry, or choose alternative paths when information is missing or incomplete.
Relevance AI agents are not scripted workflows. They are closer to digital roles that can reason, act, and collaborate.
From single agents to AI workforces
The session emphasized that meaningful scale comes from combining multiple specialized agents rather than building one monolithic agent.
- Account research agents focused on companies.
- Prospect research agents focused on individuals.
- Email specialists trained only on writing and tone.
- CRM agents limited to data hygiene and updates.
These agents can be connected in a workforce canvas, where one agent delegates tasks to others based on need. Responsibility typically sits with the agent that produces the final output.
Orchestration and delegation patterns
Workforces follow a practical orchestration model.
- A primary agent receives the task.
- Sub-agents are invoked only when required.
- Each agent operates within strict tool permissions.
This mirrors human team dynamics and reduces complexity. Changes to one role, such as email tone, can be made in a single agent and propagate across workflows.
Practical demonstrations shown
Several live demonstrations illustrated how this model works in practice.
- Automatic company research triggered by CRM events.
- Inbound email replies drafted using research and prior context.
- Outbound sales emails generated after prospect qualification.
- Content workflows that expand a single idea into posts, scripts, and assets.
Failures in individual steps did not halt the entire process. Agents continued where possible and surfaced issues for human review.
Human oversight and guardrails
The session stressed that AI workforces are not fully unsupervised systems.
- Restricted tool permissions per agent.
- Explicit instructions on what actions are allowed.
- Escalation rules for errors or uncertainty.
This approach balances autonomy with control and reduces operational risk.
Privacy and compliance considerations
Relevance AI operates with region-based data handling and enterprise-grade compliance. Language models are hosted in a way that prevents training on customer data.
When integrating third-party tools such as CRMs or email systems, responsibility remains with the user to align with those platforms’ policies.
Why this matters
The session illustrated a broader shift in how AI is applied in organizations. Instead of isolated prompts or brittle automations, teams can model their internal roles as software-native agents.
This makes AI systems easier to maintain, easier to audit, and more aligned with real business processes.
FAQ
What is an AI workforce?
An AI workforce is a coordinated set of specialized agents that collaborate to complete multi-step tasks autonomously.
Do AI agents replace automation tools?
No. Agents complement automation tools and can be triggered by or integrated with them depending on the use case.
Is this approach suitable for non-technical teams?
Yes. Agents are configured using natural language instructions and predefined tools rather than code.
Learn more
Some links may be affiliate links. This helps support the site at no additional cost and does not influence the content or reviews.