Glossary

A professional regulated by CVM who acts independently, with no employment tie to a financial institution, in the distribution of investment products to end clients. The regulation requires certified qualification, registration with CVM and operation linked to a responsible broker or distributor.

In markets such as Brazil, the AAI is the central figure of the decentralized distribution model — responsible for the direct relationship with the investor and for the ongoing fit of the portfolio against the declared suitability profile.

In PRÆTOR's context: the platform is configurable to the operational flow of the independent advisor — briefings, follow-up, WhatsApp and activity logging — with data isolation per client book.

In an AI context, an autonomous agent is a software system capable of perceiving its environment, making decisions and executing actions to reach defined objectives — without requiring human instruction at every step. Operating autonomously does not mean operating without supervision: the degree of autonomy is configurable per step and per type of action.

Unlike the AAI (the human professional regulated by CVM), the autonomous AI agent operates on behalf of the platform, with delegated autonomy, an audit trail per decision and the ability to scale across multiple clients in parallel.

Why it matters in wealth management: the autonomous agent is the functional unit that makes scale without headcount possible — it executes repetitive, high-value tasks while freeing the banker for the 30% that requires human judgment.

A category of artificial-intelligence systems capable of planning sequences of actions, using external tools, maintaining memory across sessions and operating with a high degree of autonomy to reach medium-term objectives. Agentic AI distinguishes itself from conventional chatbots by acting on the environment — not just responding to isolated prompts.

In production, agentic systems combine multiple language models, data-retrieval tools, persistent memory and coordinated orchestration of subagents.

Why it matters in wealth management: agentic AI is the architectural foundation that allows the full commercial cycle of private banking to be automated — from the morning briefing to the post-meeting interaction log.

A document or structured set of information prepared ahead of a commercial interaction. In private banking, the briefing consolidates an updated portfolio position, market events relevant to the client's profile, verified suitability, open items and suggested talking points for the banker.

When produced manually, the briefing consumes between 20 and 45 minutes of the banker's work per client. When produced by autonomous agents, it is delivered automatically before each meeting or call — in seconds, with the same quality and zero margin for omission through forgetfulness.

In PRÆTOR's context: the briefing agent operates over the client graph, pulls position data in real time, cross-references the banker's agenda and delivers a readable document 40 seconds before each contact.

Automated regulatory-adherence verification applied over AI-agent outputs before any communication or recommendation reaches the client. In agentic platforms for the institutional financial market, agentic compliance filters responses against suitability rules, CVM prohibitions, the house's investment policies and Anbima restrictions — with an audit trail per blocking decision.

Agentic compliance does not replace the bank's compliance function: it is an additional, automated layer that reduces the risk of human failure in volume and cadence.

Why it matters in wealth management: local regulators require evidence of process — not just of intent. Agentic compliance produces structured logs of every check performed before a recommendation, which is decisive in any CVM or Anbima supervision.

A customer-relationship management system that, instead of merely storing and displaying data, operates actively on it through autonomous agents. An agentic CRM schedules follow-ups, drafts replies, generates briefings, updates the client graph and logs interactions without continuous manual intervention from the banker.

The central distinction from a traditional CRM: the conventional CRM waits for the banker to act; the agentic CRM acts on its own, within configured autonomy limits, and notifies the banker only when necessary.

In PRÆTOR's context: the platform is not a CRM with AI bolted on — it is an agentic CRM by architecture, where the agents are the native operational layer, not an optional module.

A structure dedicated to the full management of a family's or family group's wealth. The family office consolidates investments, succession planning, tax management, compliance and the relationship with specialized service providers — operating as the family's own investment office, with or without an internal team.

In jurisdictions such as Brazil, the single family office (exclusive to one family) is distinguished from the multi-family office (which serves multiple families under the same operational structure). CVM regulation classifies the multi-family office as a third-party asset manager, subject to specific authorization and supervision.

Why it matters in wealth management: family offices concentrate substantial wealth per family — and operate with small teams. Agentic AI allows a team of 3 to 5 professionals to manage the complexity of dozens of portfolios at institutional quality.

A network representation of the relationships, attributes and events associated with a client and their surroundings — family, companies, partners, contracted products, recorded interactions, declared preferences and relevant life events. The graph replaces flat database records with a navigable structure that reflects the real complexity of the client's wealth situation.

Autonomous agents that operate over client graphs produce contextually more precise responses than systems that access only relational tables, because they can infer relationships — for example, identifying that a withdrawal request is associated with a corporate change recorded three weeks earlier.

In PRÆTOR's context: the client graph is the platform's central memory structure. Every interaction, document, position and event is indexed in the graph and remains accessible to the agents in future interactions — regardless of team turnover.

The capability of an AI system to retain and retrieve information across distinct sessions, without context being discarded at the end of each interaction. Conventional language models lose context at the end of each conversation; systems with persistent memory maintain a structured repository of knowledge that feeds subsequent interactions.

In wealth management, persistent memory means the agent knows the client's full history — communication preferences, prior meetings, completed operations, open items — not just the last message received.

In PRÆTOR's context: persistent memory is what distinguishes the platform from a generic AI assistant. The system accumulates institutional knowledge about each client over time — and that knowledge survives changes of banker, of system and absences of months between interactions.

Software architecture in which multiple institutional clients (tenants) operate over the same base infrastructure, with complete data and configuration isolation between them. Each tenant sees only its own data, rules and flows — with no access to information of others.

For banks with private-banking, in-house family-office and asset-management divisions under the same legal entity, multi-tenant architecture ensures each unit operates with its own identity and segregated data, even sharing infrastructure and platform license.

Why it matters in wealth management: regulators and compliance functions require information segregation between business units. Multi-tenant makes this enforceable by architecture — not by internal policy, which is auditable but breakable.

An AI agent responsible for coordinating the execution of multiple specialized subagents, distributing tasks, managing dependencies between stages and consolidating results into a coherent output. The orchestrator does not execute the tasks directly — it defines the execution plan, delegates to subagents and synthesizes the outputs.

In agentic wealth-management platforms, the orchestrator may, for example, receive a WhatsApp message from a client and trigger in parallel: the intent subagent, the position subagent and the compliance subagent — before composing the final response.

In PRÆTOR's context: the orchestrator is the coordination layer that ensures client responses are produced completely and verified, even when they involve multiple data sources and business rules simultaneously.

The wealth-management segment serving high-net-worth clients, typically with investable financial assets above an institutional threshold — which varies by institution. Private banking offers access to exclusive products, bespoke portfolio structuring, wealth planning and personalized service from dedicated bankers.

In markets such as Brazil, the largest banks operate private-banking divisions with hundreds of bankers and substantial assets under management. The segment is regulated by Bacen, with products subject to CVM supervision and Anbima rulebooks.

Why it matters: institutional private banking operates at scale — but with the complexity of personalized service. It is precisely in that tension between scale and personalization that agentic AI generates the highest operational return.

An AI agent specialized in a specific task within a larger orchestration. Subagents receive instructions from the orchestrator, execute their delimited function and return results — operating in parallel or in sequence, depending on the architecture of the flow.

In agentic platforms for wealth management, different subagents may be responsible for tasks such as briefing generation, suitability verification, drafting a WhatsApp reply, updating the client graph or identifying portfolio opportunities.

Why it matters in wealth management: specialization by subagent allows each component to be calibrated, audited and replaced independently — which is decisive for financial operations that require traceability per step.

Automated verification of a financial product's fit to the investor's risk profile and declared objectives, as required by CVM regulation (Resolution CVM 30). In agentic systems, suitability is verified in real time before any recommendation — not as a manual step by the banker, but as an automatic layer of the agent's flow.

Automated suitability does not eliminate the banker's responsibility for the final recommendation: it ensures that the agent does not compose or forward suggestions incompatible with the client's recorded profile, reducing the risk of regulatory exposure through omission.

In PRÆTOR's context: every agent output containing financial content passes through suitability verification before delivery to the banker. The verification result — positive or blocked — is recorded in a log with timestamp and justification.

Architecture in which WhatsApp is treated as the primary operational channel — not as an ancillary integration. A WhatsApp-native platform receives, interprets, processes and replies to messages directly through agents, without requiring the banker to manually transfer context to another system.

In non-native platforms, WhatsApp is monitored by a person who copies the relevant content into the CRM. In WhatsApp-native platforms, the agent reads the message, identifies the intent, consults the client graph, verifies compliance and drafts a reply — all automatically, with the banker approving or editing before sending.

Why it matters in wealth management: in markets where WhatsApp is the dominant channel between banker and private-banking client, an architecture that does not operate natively over it loses the richest context of the relationship — and forces the banker to duplicate work manually.