AI Governance · Human Review · Audit Trail

Automate customer support with control built in.

FortiAgent resolves customer support queries using grounded knowledge and live connector data. FortiVault governs every decision with an AI Trust Score and automation gating policy.

Reduced support workload, improved response consistency, safer automation expansion, and full oversight of every AI decision.

Built for enterprises that need AI customer support they can measure, audit, and safely scale.

Start free trialRequest Demo

14-day free trial · No credit card required · 1 business day setup

AI governance decision flow: incoming support requests evaluated by AI Trust Score, routed through policy thresholds to human review or automated actions, with connector sync and full audit logging

The Problem

Why Enterprise AI Customer Support Needs a Governance Layer

Most AI customer support platforms optimise for resolution rate. None of them tell you which resolutions are accurate, which categories are safe to automate, or when a response should have gone to a human instead. Resolution rate without governance is a vanity metric.

  • AI agents resolve queries at scale — but teams have no visibility into which resolutions are accurate
  • There is no threshold that tells AI when it is safe to automate a category vs. when to escalate
  • Errors in billing, returns, and account queries reach customers before anyone on the team notices
  • Automation expands informally — not gated by measured accuracy or policy thresholds

Enterprise AI customer support requires an execution layer and a governance layer — built to work together from day one, not bolted on after an incident.

The Platform

Two layers. One governed support stack.

Most enterprises deploy an AI agent without a governance layer. The result: automation that expands into categories where AI accuracy has not been validated, and errors that reach customers before anyone on the team knows the AI was wrong.

Layer 1 — FortiAgentExecution
  • AI support execution — grounded knowledge and live connector data
  • Customer interactions via embeddable widget or connected channel
  • Operational workflows and multi-step procedures
  • Escalation handoff to human agents with full transcript
Learn about FortiAgent
Layer 2 — FortiVaultGovernance
  • Trust scoring and AI accuracy measurement
  • Risk detection and human review enforcement
  • Automation governance policies per category
  • Full audit trail for every AI decision

Customers

FortiAgent

AI Execution Layer · knowledge · connectors · widget

FortiVault

Governance & Oversight Layer

Human Support Team

When escalation is required

FortiAgent handles execution. FortiVault governs every decision.

Positioning

This is not another AI chatbot.

There are meaningful differences between FortiVault and FortiAgent and the AI chatbot and copilot tools most teams have already evaluated.

Generic AI chatbots

  • —Answer from model training data — can hallucinate your policies and product details
  • —No category-level accuracy measurement or automation gating
  • —Automation expands by default — no objective threshold to check against
  • —No per-decision audit trail when something goes wrong

AI copilots

  • —Surface suggestions to human agents — useful, but not automation governance
  • —Human-in-the-loop by design — does not solve the "when is automation safe?" question
  • —No mechanism to gate AI autonomy by measured category accuracy
  • —Does not reduce human review overhead — it is human review

FortiVault + FortiAgent

  • FortiAgent answers only from your configured knowledge — no training data leakthrough
  • FortiVault measures accuracy per category and gates automation when trust isn't proven
  • Human review is a structured step enforced by policy — not a fallback after an incident
  • Every AI decision is logged and traceable, per response, before and after any override

End-to-End Flow

How the layers work together.

From customer message to governed response. Every step is measured, logged, and auditable.

1FortiAgent

Customer sends a message

Via the FortiAgent widget or connected channel.

2FortiAgent

FortiAgent retrieves knowledge and data

Knowledge sources searched. Connector APIs called if live data is needed.

3FortiAgent

FortiAgent drafts a response

Grounded in your knowledge and live connector results.

4FortiVault

FortiVault evaluates trust and policy

Trust score computed. Automation policy checked for this category.

5FortiVault

Decision: auto-send, review, or escalate

If accuracy is proven, the response is sent. If not, it enters human review.

6FortiVault

Corrections feed the improvement loop

Human overrides and audit data improve accuracy and expand safe automation over time.

Example: Billing Automation Policy

If FortiAgent trust score ≥ 90

AND billing category accuracy ≥ 92%

→FortiAgent resolves billing query automatically
→Otherwise: human review required before sending

Refunds, account changes, and other sensitive workflows require proven accuracy thresholds before FortiAgent can act without oversight.

Observability

AI Trust Score: Know When AI Is Ready to Automate

FortiVault computes an AI Trust Score across every support category — updated continuously from response accuracy, human override rate, and connector call outcomes. When the score crosses your policy threshold, automation is enabled. When it drops, human review kicks in automatically.

FortiVault · AI Trust Dashboard

AI Trust Score

82/100

↑ vs last week

AI Accuracy

87%

↑ vs last week

Automation Rate

34%

↑ vs last week

Human Override Rate

18%

↓ vs last week

Weekly accuracy trend

Last 8 weeks
W1W2W3W4W5W6W7W8

Example FortiVault AI Trust Dashboard from a calibrated support AI deployment.

See it live in the demo

Automation Governance

Automation Gating: AI Earns the Right to Automate

Most teams either automate too early — before AI accuracy in a category is validated — or never automate at all because they have no objective threshold to trust. Both outcomes are expensive.

FortiVault's Automation Gating solves this.

Each support category — billing, returns, login, technical — has its own AI Trust Score threshold. When FortiAgent's measured accuracy meets the threshold, automation is enabled for that category. When accuracy drops below it, every response requires human review before sending.

AI Automation Gate

FortiVault acts as the automation control layer. Policies ensure FortiAgent can only resolve tickets automatically when performance thresholds are met.

Example Policy

If AI Trust Score ≥ 85

AND category accuracy ≥ 90%

→FortiAgent resolves automatically

Otherwise

→Human review required

This ensures automation expands safely without exposing customers to AI errors.

FortiVault · Automation Policy Status

Billing Questions

Trust Score: 91

ENABLED

Login Issues

Trust Score: 84

HUMAN REVIEW REQUIRED

Technical Bugs

Trust Score: 72

DISABLED

FortiVault governs FortiAgent automation — category by category, based on measured trust.

FortiAgent

Connected to the systems support teams already use.

FortiAgent retrieves live data from your existing APIs — order history, account status, billing state, shipping tracking — and includes it in grounded, accurate responses.

Commerce

Shopify, WooCommerce, BigCommerce

Billing

Stripe, Chargebee, GoCardless

Helpdesk

Zendesk, Freshdesk, Intercom

Logistics

DHL, FedEx, Royal Mail, Evri

CRM

Salesforce, HubSpot, Zoho

Identity

Auth0, Okta, internal SSO

See all FortiAgent connectorsEstimate your AI support ROI

FortiVault

AI Trust Score. Automation Gating. Full Audit Trail.

01

AI Trust Score

A single, continuously updated score that reflects FortiAgent's measured accuracy across all support categories. The Trust Score determines when automation is safe — not guesswork, not manual policy review.

02

Automation Gating

FortiVault enforces automation policy category by category. Billing queries, returns, account changes — each has its own threshold. When AI accuracy is below the gate, human review is mandatory before the response is sent.

03

Full Audit Trail

Every AI decision is logged — knowledge source retrieved, connector called, guidance rule applied, and outcome recorded. Every response is traceable. No black-box behaviour.

Outcomes

AI Customer Support You Can Measure, Audit, and Scale

  • Expand automation safely — gated by AI Trust Score, not guesswork
  • Detect inaccurate AI responses before they reach customers
  • Audit every decision: knowledge source, connector call, rule applied
  • Set independent accuracy thresholds for billing, returns, and account queries
  • Know exactly when AI is ready to automate a category — and when it is not
Calculate AI support savingsAI governance deep diveHow we compare

AI Accuracy

Know exactly how often FortiAgent is correct, where it fails, and what it costs in customer experience.

AI Oversight

Enforce human review on high-risk interactions. Full audit trail for every AI decision FortiAgent makes.

Safe Automation

Expand FortiAgent automation only where accuracy supports it. Measurable, auditable, and independently reversible.

Who FortiVault Is For

Built for enterprise teams deploying AI in high-stakes support.

FortiVault and FortiAgent are not the right fit for every team. They are built for a specific requirement: AI support automation that must be measured, gated, and auditable before it scales.

FortiVault is built for you if

  • You handle billing, refunds, or account changes at scale

    These are high-stakes query types where AI errors have direct customer and revenue consequences. They require higher accuracy thresholds and stricter oversight than general FAQ deflection.

  • You operate in financial services, SaaS, or regulated e-commerce

    Compliance, legal, and audit stakeholders are asking what the AI decided, when, and why. An audit trail is not optional in your environment.

  • You've deployed AI support but aren't confident in when it's safe to scale automation

    The team knows the AI is handling some queries incorrectly. There's no objective mechanism to identify which categories are ready for automation and which aren't.

  • Human review needs to be a structured, policy-driven step — not a fallback

    You want human review enforced by the system when AI accuracy is below threshold, not triggered manually after something breaks.

FortiVault is probably not the right fit if

  • You need omnichannel reach across email, chat, and in-product surfaces natively

    Intercom and Zendesk have deep omnichannel footprint that FortiVault and FortiAgent do not replicate out of the box.

  • Your primary requirement is maximising resolution rate with no governance constraints

    If expanding automation as fast as possible is the goal and accuracy-by-category is not a concern, Fin AI or Forethought are stronger fits.

  • You are deploying a simple FAQ chatbot with no high-stakes query types

    Governance infrastructure is proportionate to risk. If your support queries are informational and low-stakes, the overhead of category-level gating exceeds the value.

Not sure if FortiVault is the right fit? The demo is the clearest way to find out. It runs on synthetic data — no production access required — and shows every governance layer in a realistic support environment.

Request access to the demo

Industry-specific guidance

Financial ServicesSaaSEcommerce

Deploy AI Support With Governance Built In

FortiAgent executes. FortiVault governs. Together they form a support stack you can measure, audit, and scale safely.

14-day free trial · No credit card required · 1 business day setup

Start free trialRequest DemoDownload Whitepaper