Safety System

5 independent safety layers. Zero surprises.

Every AI action passes through five independent control steps before it reaches your campaigns. Nothing falls through the cracks. You always have full control.

14 days free · No credit card · All safety layers active

Layer 1: Budget protection
Layer 2: Change rate
Layer 3: Rule engine
Layer 4: Approval queue
Layer 5: Audit log

Why safety isn't optional

AI managing your ad budget needs clear boundaries. Without them, you're playing Russian roulette with your money.

AI without control is a risk

Other tools give AI free rein. No review, no rules, no emergency brake. A single wrong decision can cost thousands.

Budgets can spiral out of control

Without budget limits, an AI can raise bids uncontrollably. One morning you wake up to ad spend three times your daily budget.

Wrong changes are costly

Paused campaigns at the wrong time, removed keywords that convert, or bid changes that crush your ROI. Without a safety net, there's nothing to undo.

5 layers between AI and your campaigns

Every change passes through all five layers in order. If a single layer says stop, the change is stopped.

Layer 1

Budget Protection

Monitors budget limits in real time. Sets maximum daily and weekly change. Your ad budget can never be exceeded.

Example

Scenario: An agent wants to increase the budget by 40%. Budget protection stops the change because the max limit is 20% per day. You get a notification instead.

Configure max percentage increase per day and week. Set hard budget caps per campaign. All violations are logged and reported.

Layer 2

Change Rate

Limits the number of changes per day and week. Prevents the AI from making too many changes at once, which can destabilize campaign performance.

Example

Scenario: The agent wants to change bids on 15 keywords simultaneously. The change rate allows max 5 per day. The remaining 10 are queued for tomorrow.

Set max number of actions per day, per week and per campaign. New campaigns automatically have stricter limits during the learning period.

Layer 3

Rule Engine

Custom rules per campaign and client. Define business rules that must always be respected — regardless of what the AI suggests.

Example

Scenario: You set the rule 'Never bid above 50 kr' on a campaign. The agent wants to bid 65 kr. The rule engine blocks the change automatically.

Create rules for bid limits, scheduling, pausing and more. Rules can be set per campaign, per client or globally. Rule violations are always logged.

Layer 4

Approval Queue

All changes that pass the first three layers end up in an approval queue. You review and approve before anything is executed.

Example

Scenario: The agent suggests pausing a campaign with low ROI. The change appears in your approval queue with full reasoning. You decide.

Optional: at autonomy level 'Moderate' or 'Autonomous' routine changes are executed automatically. Major decisions always require approval.

Layer 5

Audit Log

Every action is logged with timestamp, before/after data, AI reasoning and results. Full traceability with undo function.

Example

Scenario: A bid proposal was approved yesterday but resulted in higher CPC. You open the log, see before/after data, and click 'Undo' to restore.

Complete audit history. Filter by agent, campaign or action type. Undo any change with one click. Export logs as CSV.

Budget Protection
Change Rate
Rule Engine
Approval Queue
Audit Log
Executed

You determine the degree of automation

Start cautiously. Increase autonomy as your confidence grows. It's not an all-or-nothing question.

Suggest

Recommended start

All changes require your approval. The AI analyzes and recommends — you decide.

Automatic

Nothing

Requires approval

All actions

Moderate

Routine changes like negative keywords and minor bid adjustments happen automatically. Major decisions require approval.

Automatic

Negative keywords, minor bid adjustments

Requires approval

Budget, strategy, pausing

Autonomous

Full automation with all 5 safety layers active. The AI handles the daily work. You monitor on your schedule.

Automatic

Most things — with safety layers

Requires approval

Major strategy and budget changes

Suggest
Moderate
Autonomous

Most start with "Suggest" and increase gradually. Change level at any time.

What happens without safety layers?

Compare: AI without a safety net vs Dirigent's 5-layer safety system.

Aspect
Without protection
With Dirigent
Budget control
AI can increase budget indefinitely
Max daily/weekly change
Change frequency
Hundreds of changes per day
Controlled pace, queued changes
Business rules
AI decides entirely on its own
Your rules enforced automatically
Review
None — changes happen directly
Approval queue with full reasoning
Traceability
No history, no undo button
Complete log, undo with one click
Disaster protection
You discover problems too late
5 layers stop problems before they happen

Try the safety system free for 14 days

All 5 safety layers active from day one. See exactly what the AI does, why it does it, and stop anything you don't like.

No credit card · All features included · Cancel anytime