OptiOps AI Platform - Auto-Remediation Demo

Learning Mode

IT Operations Auto-Remediation

check_circleAction Completed Successfully

• Background job service restarted

• CPU normalized to 23% (was 94%)

• Server DB-07 healthy

• Completed at 2:47 PM by Ken Hung3

3 selected

monitor_heart

Monitor

Check 47 servers

0.3s

Detect

Pattern matching

1.0s

psychology

Analyze

Root cause analysis

1.2s

shield

Approval Gate

Human review

build

Remediate

Execute solution

2.1s

notifications

Notify

Send alerts

0.4s

Node Configuration

Click on a workflow node to view and edit its configuration

Live Execution Trace

Click "Test Workflow" to begin execution...

Approval Policy Configuration

checkAuto-Execute (No approval needed)

✓ Single server, confidence >85%

✓ Non-production environments

✓ Low-risk actions (restart monitoring)

warningRequire Approval (You must review)

⚠️ Multiple servers affected

⚠️ Confidence <85%

⚠️ Production database servers

priority_highAlways Escalate (Senior engineer required)

🚨 Actions affecting >10 servers

🚨 Downtime >5 minutes

🚨 Customer-facing services

Progressive Autonomy Control

ManualFully Autonomous

Current position Preview:
• 73% of actions would auto-execute
• 22% would require your approval
• 5% would escalate to senior engineer

Monitor Configuration

dnsMonitored Resources

• Loading...

scheduleCheck Interval

⏱ Loading...

📊 Loading...

sensorsAlert Thresholds

⚠️ Loading...

Detection Configuration

patternDetection Patterns

✓ Loading...

model_trainingMachine Learning Models

🤖 Loading...

speedConfidence Scoring

Analysis Configuration

troubleshootAnalysis Method

🔍 Loading...

library_booksKnowledge Sources

📚 Loading...

balanceImpact Assessment

Action Configuration

constructionAvailable Actions

🔧 Loading...

safety_checkSafety Measures

✓ Loading...

undoRollback Policy

⏱ Loading...

Notification Configuration

sendNotification Channels

📧 Loading...

💬 Loading...

📱 Loading...

📊 Loading...

🎫 Loading...

priority_highAlert Priorities

🔴 Loading...

🟠 Loading...

🟡 Loading...

🟢 Loading...

descriptionMessage Template

Orchestrated AI Agent — TL;DR — Ken Hung

Prototype 04 Enterprise Workflows · Multi-Domain Designing for Different AI Archetypes

Orchestrated AI Agent

Autonomous Agent · Orchestrated Workflow

The governance features got built. The operator experience didn't.

User Problem

Operators are responsible for systems they can't see or control.

Enterprise AI agents run continuous workflows across entire organizations — monitoring systems, routing tickets, reallocating budgets, moderating content — often executing dozens of actions per hour without direct human involvement. Every major platform now ships governance capabilities: dashboards, audit trails, compliance consoles.

But governance tooling is not the same as governance experience. A non-technical operator responsible for a system taking hundreds of autonomous actions on their behalf doesn't need more dashboards — they need an interface that makes them feel genuinely in control. The features got built. The operator experience of actually feeling in control didn't.

Design Principles

Five principles of Orchestrated AI Agent UX

Pipeline Visibility

When a system is executing dozens of actions across multiple domains, operators need to understand its state without reading a log. A visual workflow canvas communicates system status spatially — green nodes completed, amber nodes waiting, the Approval Gate pulsing when human input is required. An operator who hasn't checked in for an hour should understand the full system state in under three seconds.

Human Placement

In most enterprise AI systems, the approval step is an interruption — a notification or modal that appears outside the workflow. The better model embeds the human decision point structurally inside the workflow as a literal node in the sequence. Human judgment isn't an override of the process — it's a step that was always part of it. That distinction changes the operator's mental model entirely.

Configurable Autonomy

A binary on/off switch for AI autonomy is not governance — it's avoidance. A three-tier approval policy (Auto-Execute, Require Approval, Always Escalate) gives operators a graduated spectrum, with every threshold defined by them and visible in the interface. A progressive autonomy slider makes calibration tangible: operators see the projected operational impact before committing.

Audit Transparency

Transparency at enterprise scale must serve two audiences: non-technical operators who need ambient state awareness, and engineers who need step-by-step log detail. A visual canvas serves the first. A live execution trace with timestamps and source attribution serves the second. The approval modal adds a third layer — a confidence formula broken down line by line, with a full reasoning chain available on demand.

Framework Portability

Enterprise organizations run dozens of AI workflows across IT, support, marketing, and compliance — each with different risk profiles and approval triggers. The governance UX must be consistent enough for operators to transfer skills across domains, but flexible enough to reflect each domain's actual risk profile. When the structural UX stays constant and only the content layer changes, operators who learn one domain can immediately operate the next.

Design Solution

One framework, four organizational contexts

The prototype ships with four ready-made use cases. The same six-step workflow — Monitor → Detect → Analyze → Gate → Act → Notify — handles every domain. The governance UX stays constant. Only the content layer changes per domain.

IT Operations Auto-Remediation

Monitors 47 servers. Triggers gate when: confidence <85% or production DB affected.

Medium Risk · 5 min review window

Customer Support Ticket Routing

Reads email, chat, in-app tickets. Triggers gate when: legal language or enterprise account ($50K+).

High Risk · Immediate review

Marketing Budget Auto-Allocation

Tracks campaigns across platforms. Triggers gate when: budget move >$10K or confidence <80%.

Medium Risk · 24h review window

✦

Content Moderation Auto-Review

Scans posts and flagged messages. Triggers gate when: confidence <70% or clean user history.

Medium Risk · 4h review window

The prototype also includes a visual workflow canvas with drag-and-drop nodes, animated execution state, and color-coded node status; a live execution trace panel running in parallel with the canvas; an approval modal with a line-by-line confidence formula and expandable reasoning chain; and a three-tier approval policy settings panel with a progressive autonomy slider showing projected split percentages before any policy is committed.

Key Insight

"At enterprise scale, agentic AI governance isn't a modal dialog — it's an operating model. The UX must make policy visible, adjustable, and auditable by the people responsible for the outcomes."