Salesforce Agentforce 360 for Public Sector AP flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Atlas Reasoning Engine

The reasoning layer inside Agentforce that interprets a request, selects the matching topic, plans and sequences actions, and decides when to respond or hand off to a person.

Autonomous agent

An Agentforce agent that does not wait for a live conversation: it is triggered by events or schedules to complete background work such as record updates and follow-ups.

Agentforce Service Agent

A customer-facing Agentforce agent deployed on messaging channels and Experience Cloud sites that answers questions and runs actions, escalating to human representatives when needed.

Agentic enterprise

An operating model in which AI agents work alongside people, reasoning over trusted enterprise data and executing multi-step processes rather than only suggesting next steps.

Einstein Agent user

The dedicated user license and execution context under which an agent's actions run; its assigned permissions and sharing determine which records and operations the agent can reach.

Human handoff

Escalation in which the agent transfers a conversation or work item to a person through Omni-Channel, passing context so the customer does not repeat information.

Topic classification description

The text the reasoning engine compares a request against to choose a topic; it should carry the distinct vocabulary and typical requests of that one job.

Topic scope

A topic field that states what the agent should and should not handle, so out-of-scope requests are refused or routed instead of answered speculatively.

Agent action

A callable capability the agent invokes from a topic, such as a Flow, Apex class, prompt template, standard action, or external API, each with defined inputs and outputs.

Agent Script

A code-like authoring layer in Agentforce Builder that uses typed variables and conditional logic so an agent's steps follow deterministic instructions rather than pure model planning.

Prompt template action

A reusable, versioned prompt with merge fields that the agent populates from conversation context or record data before sending it to the language model.

Agent-level instructions

Global guidance applied to every topic the agent handles, covering persona, tone, formatting conventions, and universal guardrails that no topic may override.

Retrieval augmented generation

A grounding technique that first retrieves relevant records or document chunks, then inserts them into the prompt so the model answers from enterprise data rather than training memory.

Data 360 data stream

A connector-based ingestion of source data into Data 360 that lands records in data lake objects before they are mapped into the data model.

Vector search index

An index of embedded document chunks supporting semantic similarity search, letting an agent retrieve passages matching the meaning of a question rather than its keywords.

Einstein Data Library

The Data 360 capability that ingests unstructured files such as PDFs, splits and embeds them, and exposes them for retrieval-grounded agent answers.

Identity resolution

Data 360 matching rules that reconcile records from multiple sources into one unified profile, giving an agent a single view of a constituent or household.

Data space

A logical partition in Data 360 scoping which data and which users belong together, used to segregate agencies or programs and keep retrieval permission-aware.

Calculated insight

A Data 360 metric defined with SQL-like expressions over data model objects, refreshed on a schedule and reusable in segments and grounded agent answers.

Government Cloud Plus

Salesforce's highest-assurance US government environment, supporting FedRAMP High and DoD Impact Levels 2, 4, and 5 for sensitive public sector workloads.

Licensing and Permitting

The Public Sector Solutions module that models permit types, applications, fees, and reviews so an agency can accept and adjudicate permits digitally.

Inspection Management

The Public Sector Solutions module for scheduling inspections, capturing findings on mobile, and generating violations and follow-up work from the results.

Grant Management

The Public Sector Solutions module covering the grant lifecycle: application intake, review and scoring, award, disbursement, and reporting obligations.

OmniStudio

A declarative toolset of OmniScripts, FlexCards, and Integration Procedures used in Public Sector Solutions to build guided multi-step constituent interactions and data lookups.

311 service request

A non-emergency constituent request such as a pothole or sanitation issue, modeled as a case and routed to the responsible department with status tracking.

Business Rules Engine

A Public Sector Solutions engine that evaluates declarative eligibility and determination rules and returns decisions that flows and agents can consume.

Einstein Trust Layer

The control layer between Salesforce and the language model that masks sensitive data, blocks provider retention, detects toxicity, and logs prompts for audit.

Zero data retention

A trust setting under which the model provider stores neither prompts nor completions and cannot train on customer data; bring-your-own-model deployments shift that responsibility to the customer.

Dynamic grounding

Trust Layer retrieval of only the records and passages the running user is permitted to see, injected into the prompt at request time.

Agentforce Testing Center

A tool for defining test cases with expected topics and actions, running them in batches, and comparing results before an agent version is activated.

Prompt injection

Input crafted to make an agent ignore its instructions or leak data, mitigated by input scanning and by keeping secrets out of agent and topic instructions.

Agent audit trail

Trust Layer records of prompts, masked payloads, and model responses kept for review, used as evidence in incident investigation and compliance checks.