AI Governance & Trust Center

Your CRM data never leaves your org.
Every AI action is fully audited.

Most AI tools mean a parallel security model — separate audit logs, separate FLS, separate data pipeline. REDA AI runs inside your Salesforce org and inherits your existing FLS, sharing rules, and profiles automatically. Salesforce AI governance built in by design. ISO 27001:2022 and SOC 2 Type II certified.

ISO 27001:2022 Certified
ISO 9001:2015 Certified
GDPR Compliant
SOC 2 Type II
100% Native Salesforce
SOC 2 Type II Certified
The Native Claim, Defined

What "Salesforce-Native" Actually Means Here

Native is a strong word. We use it precisely. Here are the five architectural commitments behind it — so technical buyers can verify the claim, not just read it.

Managed Package Inside Your Org
REDA AI is installed as a managed package directly into your Salesforce org. Apex, LWC, custom objects, and admin UI all live inside your Salesforce org. There is no separate REDA AI application server holding state.
Configuration, Logs, Audit in Salesforce
Agent definitions, prompts, conversation transcripts, audit trails, incident logs, and credit usage are stored as standard Salesforce objects in your org. You query them with SOQL, govern them with sharing rules, and back them up with your existing Salesforce data lifecycle.
Security Enforced Before Context Assembly
FLS, sharing rules, profiles, and object permissions are checked before any record context is loaded for an AI interaction. The agent only sees what the running user is allowed to see — no privilege escalation, no shadow read paths.
Inference via Approved Model Provider APIs
Reasoning happens at an approved AI model provider (the LLM does not run inside Salesforce). REDA AI sends the minimum prompt context required for the request and is governed by the provider's data-processing terms. No customer fine-tuning corpus, no model training on your data.
No REDA-Side CRM Database
REDA AI does not maintain a separate database holding copies of your Salesforce records. There is nothing to compromise outside your org boundary. You own the data, the schema, and the audit trail in the same way you own any other Salesforce object.
Verifiable, Not Just Claimed
Every commitment above is testable in your trial org: review the package contents in Setup, query the audit objects with SOQL, inspect the FLS-aware tool execution in the Incident Log. The architecture is the proof.
Data Residency

Your CRM Data Stays in Salesforce. Inference Context Is Tightly Scoped.

REDA AI is a Salesforce-native managed package. There is no REDA-side server holding CRM records, no third-party database storing your data, and no pipeline replicating your org. Here's what stays in your org and what is sent — minimally and per provider terms — to the AI model for each interaction.

What Stays Inside Your Org Conversations, transcripts, audit logs, agent configurations, incident records, credit usage — all stored as standard Salesforce objects.
What Is Sent for AI Inference Only the assembled context you configure: agent persona + prompt instructions + configured context fields + user message. No raw CRM records. No PII unless you choose to include it.
What Is NOT Stored Externally Conversation transcripts, field values, user data, record IDs, or any Salesforce record data. The AI model provider does not persist your data.
Not Used for Training Your data is never used for model training, fine-tuning, or any learning shared with other customers. Each API call is stateless and isolated.
Lifecycle of an AI Interaction
1
User Sends Message Message entered inside Salesforce. Stays in your org.
↓
2
Context Assembled with FLS Only configured fields loaded. Field-level security applied. Unconfigured fields never touched.
↓
3
Inference Request Sent Assembled context sent to AI model. No raw CRM records. No persistence at AI service.
↓
4
AI Response Returned Response displayed in Salesforce. Write actions require user confirmation via Action Cards.
↓
5
Conversation Stored in Your Org Full transcript saved as Salesforce records — user, timestamp, agent, credits, record context.
Zero External Data Storage Raw Salesforce data never exits your org boundary
Access Control

Your Salesforce Security Model — Fully Inherited

REDA AI doesn't create a separate access model. It inherits the one you already manage — profiles, permission sets, sharing rules, and field-level security. The AI sees exactly what the user sees. Nothing more.

Field-Level Security

If a user cannot see a field in Salesforce, the agent cannot see it either. FLS is enforced on every tool call — query_records, get_record, context fields, and all action tools.

Result: Sensitive fields (SSNs, financials, health data) stay invisible to the AI unless explicitly configured AND permitted by FLS.

Object-Level Permissions

Agents only access objects the user's profile permits. If a user's profile doesn't grant access to an object, neither does the AI agent. No privilege escalation — ever.

Result: Custom objects, standard objects, and sensitive data objects all respect existing profile-level access controls.

Record Sharing

Org-wide defaults, role hierarchy, sharing rules, and manual shares are all respected. The AI sees exactly what the logged-in user sees — no more, no less.

Result: A sales rep's agent only sees their own opportunities. A manager's agent sees the team's records. Sharing model fully honoured.

REDA AI Permission Sets

Four purpose-built permission sets — Admin, User, Community User, and External Guest User — each granting only the minimum access needed for their role.

Result: Admins configure agents; users interact with them; community and guest users get heavily restricted access. Principle of least privilege.

Credit Management & Optimization

Org/month credit model with built-in management tools — set limits per user, per session, and per hour. Real-time usage dashboards, efficiency ratings, and optimization recommendations give administrators complete transparency and predictability.

Result: Full cost control with built-in tools to optimize usage. No runaway AI spending. Complete transparency at org, user, and agent level.
Per-Channel Security

Each Deployment Channel. Its Own Security Posture.

REDA AI deploys across 8 channels — each with a distinct authentication model, data access scope, and security controls. Your compliance team can evaluate each independently.

Internal Channel

Lightning record pages, utility bar
Auth
Salesforce session (existing user login)
Data Access
Full FLS, sharing rules, profile permissions
Context
Record page context fields loaded automatically
Actions
Action Confirmation Cards required for all write operations
Full Salesforce Security

External Widget

Public website, landing pages
Auth
Guest User profile (no login required)
Data Access
Guest User permissions only (heavily restricted)
Security
Allowed Origins whitelist, API key encryption, daily credit caps
Actions
Conservative tool assignment recommended (read-only)
Restricted Access

Community / Experience Cloud

Customer portals, partner portals
Auth
Portal user credentials (authenticated login)
Data Access
Portal user's sharing model (own records + shared)
Security
Record-level isolation enforced, no cross-account access
Actions
Conservative tool assignment, restricted action scope
Portal-Scoped Security

API / Flow Channel

Salesforce Flows, scheduled automation
Auth
Admin-defined context — User Mode, System with Sharing, or System without Sharing
Data Access
Determined by Flow's configured run context, not fixed to system permissions
Security
No UI exposure, results consumed programmatically
Actions
Auto-execute available for unattended automation
Admin-Configured Access
Audit & Monitoring

Complete Visibility Into Every AI Action

Every AI interaction is logged at the record level — giving administrators and compliance teams the granular traceability enterprise governance demands.

Conversation Audit Log

What is captured on every AI interaction
User IdentityWhich Salesforce user initiated the conversation
TimestampExact date and time of every message exchange
Record ContextWhich Salesforce record the conversation was tied to
Channel & SessionWhich deployment channel and session ID
Credit UsageCredits consumed per message, per session, per user
Agent & Tools UsedWhich agent handled the conversation and which tools were called

Incident Log

Automatic capture of any error or anomaly
Error ClassificationSeverity level: Critical, High, Medium, Low
Root Cause DataError type, method, and stack context — without exposing raw user data
User & Record AttributionWhich user and record were involved in the incident
Status TrackingOpen → In Review → Resolved — track remediation lifecycle
Debug with AIAI-assisted root cause analysis built directly into the incident record
Resolution NotesCapture and retain remediation history for compliance records

Credit Audit Trail

Track every unit of AI consumption
Per-Interaction LoggingEvery AI call logged with exact credit cost
Per-User BreakdownTrack consumption by individual user across all sessions
Per-Agent AnalyticsCompare credit efficiency across different agents
Efficiency RatingsOptimal / Good / Needs Review / High Cost — per conversation

Data Retention & Archival

Lifecycle management for conversation data
Configurable RetentionDefine how long conversation data is kept — per agent or org-wide
Automated ArchivalOld conversations automatically archived via batch processing
Archival LogsFull statistics tracked in AI Archival Log with event history
Compliance ExportExport conversation data for regulatory review — standard Salesforce reports
Continuous Monitoring

Every metric rolls up into a live dashboard — shipped with every deployment.

Open Incidents
47
All Time · AI-debugged
Incident Trend
Oct '25NovDecJan '26Feb
Scheduled Actions
6
Total
Active Paused Draft
Guardrail Prompt Examples
"Never share pricing information with external users" Restrict sensitive business data from the external widget channel
"Always escalate to a human when the user mentions legal action" Define escalation rules for high-risk conversations
"Do not create records on objects outside the approved list" Restrict write actions to a defined set of objects
"Do not answer questions unrelated to the customer's account" Keep conversations focused and on-topic
Configurable Guardrails

Define What the AI Cannot Do

Enterprise compliance teams need to know: can we restrict what the AI does? With REDA AI, the answer is yes — at every level. Guardrail prompts, action confirmations, and rate limiting give you full control.

Guardrail Prompt Type A dedicated prompt type for defining restrictions, escalation rules, and boundary conditions. Written by admins in plain language.
Action Confirmation Cards Every write action (create, update, delete) displays an Action Card requiring explicit user approval before execution. Human always in control.
Credit Management & Optimization Tools Org/month credit model with built-in management tools. Configurable hourly, per-session, and per-user limits. Real-time dashboards and optimization recommendations for complete cost transparency.
Certifications & Standards

Independently Certified. Verifiably Compliant.

REDA AI is independently certified and audited to meet the standards your enterprise, legal, and procurement teams require.

🛡️

ISO 27001:2022

Information Security Management System. Demonstrates systematic, rigorous data protection across all operations.

✓ Certified
✅

ISO 9001:2015

Quality Management System. Confirms consistent, high-quality service delivery and continuous improvement processes.

✓ Certified
🔒

SOC 2 Type II

Independently audited for security, availability, and confidentiality controls over a sustained operational period.

✓ Certified
🇪🇺

GDPR Compliant

All data stays in your Salesforce org. Your existing GDPR controls apply. Full audit trail supports right of access and right to erasure.

✓ Compliant

ISO 27001:2022 and SOC 2 Type II Certified

REDA AI is built 100% natively on Salesforce and holds ISO 27001:2022, ISO 9001:2015, and SOC 2 Type II certifications, with GDPR alignment. Architecture is fully native: no external servers, no data egress.

Certified
Security FAQ

Questions Your Security & Legal Teams Will Ask

Direct answers to the questions we hear from enterprise IT, security, compliance, and procurement teams.

Data & Privacy
No raw Salesforce records are ever transmitted to the AI model. REDA AI only sends the specific context fields you explicitly configure for each agent — nothing else. Your accounts, contacts, opportunities, and other CRM records remain inside your Salesforce org at all times.
No. REDA AI uses the AI model in inference-only mode via API. Your organisation's data is not used for model training, fine-tuning, or any learning that would be shared with other customers. Each API call is stateless and isolated.
All data — every conversation, log, configuration, and audit record — is stored in standard Salesforce objects inside your own Salesforce org. REDA AI has no external database, no cloud storage, and no third-party data store.
Administrators control exactly which fields are made available as context to the AI — called Context Fields. Any field not explicitly included is never sent. Sensitive fields such as SSNs, financial details, or health information can be excluded entirely. Combined with Salesforce field-level security, you have full control over what the AI can and cannot access.
REDA AI is powered by world-class AI models managed by REDA. You don't need to manage model versions, API keys, or AI infrastructure. The model is accessed via secure API calls — no data is persisted at the AI model provider.
Inference requests are processed by the AI model provider's infrastructure. Only the assembled context you configure is sent — no raw Salesforce records. The AI model provider does not store, log, or persist any of your data beyond the duration of the API call.
Access & Permissions
No. REDA AI agents run in the context of the logged-in Salesforce user. The agent can only query and act on records that the user already has access to under your Salesforce profiles, permission sets, and sharing rules. There is no service account, elevated permission, or privilege escalation of any kind.
Access to REDA AI agents is managed through standard Salesforce permission sets and profiles. Administrators can restrict agent access to specific user groups, profiles, or individual users — using the same access model your team already manages.
Yes. Tool assignment and context field configuration control exactly which objects and fields are available to each agent. An agent can only access what has been explicitly enabled in its configuration — combined with the user's own Salesforce FLS and object permissions.
AI Governance & Controls
Each REDA AI agent is configured by your administrator with a specific set of tools, context fields, and instructions. The agent can only perform actions within its defined scope. Admins also configure Guardrail prompts to define restrictions, escalation rules, and boundary conditions. All configuration is visible through a transparent preview panel.
Incident logging captures every error automatically. The "Debug with AI" feature analyses the conversation to identify root cause. Guardrail prompts can enforce escalation rules (e.g., "escalate to human when the user mentions legal action"). Action Confirmation Cards prevent the AI from taking write actions without explicit user approval.
Yes. REDA AI uses an org/month credit model with built-in management and optimization tools. Administrators can set limits per user, per session, and per hour. Real-time usage dashboards, per-agent efficiency ratings, and optimization recommendations provide complete transparency and predictability — all tools to manage and optimize credit usage are included in every plan.
Every conversation is logged as a Salesforce record with full transcript, tools called, credits consumed, user identity, timestamp, and record context. All audit data is stored in standard Salesforce objects — queryable via reports, dashboards, or any SIEM integration you already use.
Compliance & Certifications
Yes. All data remains within your Salesforce org. No raw personal data is transmitted externally. The full audit trail supports right of access and right to erasure. As a Salesforce-native app, REDA AI also inherits Salesforce's own GDPR commitments as a data processor.
Yes. Contact legal@reda.one for a copy of our Data Processing Agreement. We support standard DPA requirements for enterprise procurement and can work with your legal team on any customisations needed.
Yes. REDA AI holds ISO 27001:2022 and ISO 9001:2015 certifications, and is SOC 2 Type II certified. GDPR alignment is additionally in place, with all data remaining inside your Salesforce org.
Yes. Conversation and incident data are standard Salesforce records — compatible with any Salesforce reporting, monitoring, or SIEM integration you already use. No additional connectors or APIs needed.
Yes. We provide security questionnaire responses, compliance documentation, and can arrange a technical security briefing for your IT and security teams. Contact our team at security@reda.one to request documentation or schedule a review call.
Talk to Our Security Team

Enterprise AI Governance & Salesforce AI Security Review

Our team is ready to answer your enterprise AI governance and Salesforce AI security questions, provide DPA and audit documentation, and schedule a dedicated technical briefing for your IT and legal teams.

Request Security Documentation Book a Security Briefing
security@reda.one  ·  legal@reda.one  ·  Response within 24 hours

Salesforce-Native AI Governance, Defined

REDA AI governance is architectural. Salesforce records, conversation logs, and audit trails stay inside the customer's Salesforce org as standard objects — there is no REDA-side database, no Data Cloud dependency, and no third-party CRM storage. REDA AI enforces Salesforce field-level security, profile permissions, and sharing rules natively, before any context is assembled, with a full audit trail on every AI interaction. The platform is ISO 27001:2022 certified and GDPR aligned, with configurable guardrails available out of the box. REDA AI is a Salesforce-native alternative for teams that want governed AI without bolting on a separate platform.

Last updated March 2026