Agent Taxonomy

Quick Definition

The MCI Agent Taxonomy is the strategic functional classification of AI agents (IAm) into four fundamental families: Acquisition, Qualification, Support, and Expansion. It establishes clear mandates, operational limits, and specific KPIs for each agent, ensuring that artificial intelligence acts with a defined purpose at every stage of the journey.

In Simple Terms

There is no point in putting a "do-it-all" AI to talk to your customer, as it will end up doing nothing well. Agent Taxonomy is the organizational chart of your digital workforce. It defines who attracts the customer, who discovers if they have the profile to buy, who answers technical questions, and who offers the next product, ensuring each robot knows exactly where its responsibility begins and ends.

Why This Concept Exists

Companies suffer from "Decision Gaps" and "Operational Amnesia" when trying to implement generic AIs. Without a taxonomy, agents become confused: a support bot tries to sell when the customer just wants a bank slip, or a sales bot doesn't know how to handle a technical complaint. Agent Taxonomy resolves operational chaos, allowing precision at scale and avoiding the waste of tokens on interactions without a business objective.

Teaching Metaphor

Imagine the kitchen of a fine dining restaurant. You don't just have generic "cooks." You have the Saucier (sauces), the Patissier (pastry), the Gardemanger (cold starters), and the Chef de Cuisine (coordination). If the pastry chef tries to prepare the meat, the dish is ruined. Agent Taxonomy defines the workstations in your conversational operation: each agent masters its technique and delivers the perfect "dish" to the next in line.

Practical Example

In a B2B Software (SaaS) company:

• Acquisition Agent (Ads/WhatsApp): Impacts the "Tourist," offering a practical guide in exchange for initial context.
• Qualification Agent (Screening): Chats with the "Explorer" to validate the Conversation Score. If the score is high, it schedules with a human or passes to the next stage.
• Support Agent (Knowledge Base): Serves the customer in the "Experience" state, resolving doubts about platform settings.
• Expansion Agent (CS/Upsell): Identifies that the customer has reached a usage limit and proactively suggests upgrading to a higher plan, focusing on the C for "Convenience."

Anti-example

An "FAQ Bot" that has a "Talk to a Consultant" button and nothing else. This is not a taxonomic agent; it is just a rigid decision tree. Agent Taxonomy requires autonomy within a mandate: the agent must be able to reason and decide the Next Best Action within its domain, rather than just repeating canned phrases.

How It Appears in Operation

• Specialization: Reduction of unnecessary human handoff, as the agent resolves what is within its scope.
• Continuity: The Support Agent reads the Bandeja de Contexto left by the Qualification Agent (no amnesia).
• Cost Efficiency: Less consumption of expensive models (LLMs) for simple acquisition tasks.
• Conversion: Increase in response speed (Time-to-Response) at each stage of the dynamic journey.

How to Apply in MCI

In MCI, Agent Taxonomy is the engine that moves the customer through the 6 Decision States.

• Acquisition Agents focus on the Trigger.
• Qualification Agents manage Exploration and Comparison.
• Support and Expansion Agents ensure Experience and loyalty. They feed the Conversational Memory and respect the 8Cs, ensuring that Trust is not broken by contradictory information. The Guardião do Ciclo monitors if the transition between these agent families is occurring without friction.

Related Metrics

• Resolution Rate by Mandate: How much the agent resolved within its scope.
• Qualified Handoff Rate: How many times the agent correctly passed the baton.
• Conversation Score Evolution: How much the agent helped evolve the customer profile.
• Cost per Interaction (per Family): Profitability of each stage of the taxonomy.

Diagnostic Questions

• Do our AI agents have clear business goals or are they just "search interfaces"?
• If a support customer shows purchase intent, does our agent know how to identify it or does it ignore the signal?
• Does the agent serving on Instagram have the same "training" and limits as the agent serving on the customer portal?
• Is there a decision hierarchy between agents to prevent overlap?

Related Terms

• IAm (Marketing Artificial Intelligence): The technological entity that assumes the taxonomy roles.
• Bandeja de Contexto: Where agents from different families exchange information to avoid amnesia.
• Guardião do Ciclo: The human or systemic role that audits if the taxonomy is being respected.
• Conversation Score: The metric that dictates when a Qualification Agent should pass the ball.

Executive Mode

For C-Level leadership, Agent Taxonomy represents the allocative efficiency of intelligence. Instead of investing in vague AI projects, you invest in functional assets with clear ROIs: agents to reduce CAC (Acquisition/Qualification) and agents to increase LTV and reduce Churn (Support/Expansion). It is the strategy for scalable growth without the proportional increase in Headcount.

Operational Mode

For managers, taxonomy facilitates training and continuous improvement. You don't "adjust the AI" generically; you adjust the "Mandate Prompt" of the Qualification Agent because it is letting bad leads through, or you calibrate the Bandeja de Contexto of the Support Agent because it doesn't know what was sold. It provides clarity on what to demand from each automation.

Technical Mode

Architecturally, Agent Taxonomy translates into Agent Swarms or Multi-Agent Architectures. Each family can run on different language models (LLMs): a smaller, faster model for Acquisition (low latency) and a more robust, critical model for Expansion/Technical Sales. System Prompts, tools/functions, and access tokens are defined for each agent based on its taxonomic domain.

Playful Mode

Imagine a relay race. The Acquisition Agent runs the first 100 meters and passes the baton (context) to the Qualification Agent. This one, in turn, runs until it delivers it to the Human Salesperson or Closing Agent. If the Acquisition runner tries to run the entire race alone, they get tired and lose the rhythm. Taxonomy ensures that each runner is rested and specialized in their stretch of the track.

Executive Summary

Agent Taxonomy is the organizational structure of the modern company guided by the MCI methodology. It ends the era of generic chatbots and inaugurates the era of Specialist Agents, transforming AI into a workforce categorized by Acquisition, Qualification, Support, and Expansion. Without taxonomy, there is chaos and amnesia; with taxonomy, there is a dynamic journey and profitable scale.