The technology

The engine that
grades itself.

Most ad tools fire rules and hope. CallRadius runs a patent-pending closed-loop engine (U.S. App. No. 64/063,539) where every engine feeds every other — and where the system has to prove each decision helps before it's allowed to keep making it.

Cadence
84×/week
Triage gate
≥85%
Autonomy
3-tier
Lead Budget Position Revenue Geo Reviews Closed Loop self-regulating
Six signals, one loop — every result becomes the next decision's input.
System architecture

How a decision actually gets made

Signals flow in, engines score them, a risk-tiered decision layer decides what the system may do alone — and every outcome flows back as new input.

SIGNALS IN Calls & messages Spend & pacing Reviews & position Geo & competitors 8 ENGINES · SCORE Lead triage · credit recovery Budget optimization Speed-to-lead Reviews · position Geo · seasonal every ~2 hours, per account DECISION LAYER Hierarchical suppression protective overrides growth Risk-tiered autonomy confidence × reversibility Outcomes ledger grades & demotes rules ACTIONS Adjust budget Dispute leads Reply & escalate Alert a human every outcome feeds back in — the closed loop
Ingest → score → decide (with guardrails) → act → measure → feed back. No step operates in isolation.
Mechanism 01

Hierarchical rule suppression

Every rule engine faces conflicting signals — one rule says "grow, demand is up," another says "pull back, ROI is slipping." Naive engines average them and oscillate.

CallRadius ranks them instead. Protective rules (overspend, poor ROI, low lead quality) pass through a boolean gate that can suppress growth rules entirely — so the system never grows spend into a losing situation.

Solves the conflicting-signals problem that makes other engines wobble
Safety always wins the tie — by design, not by luck
GROWTH RULES Demand surge +20% Win-rate high +15% PROTECTIVE RULES ROI below floor Lead quality low GATE growth suppressed Safe budget action
When a protective rule fires, it suppresses the growth rule outright — safety wins the tie.
Mechanism 02

The AI credit-recovery loop

Lead quality isn't just reported — it's closed back into the budget. Confidence-gated triage reads every lead; only classifications above an ≥85% confidence threshold auto-submit a dispute to Google.

The resulting credit rate then feeds the budget quality gate: markets and hours that generate disputable junk get less spend, and the clean ones get more. Quality and budget become one signal.

Recovers credit most advertisers never bother to claim
Ambiguous leads are held for human judgment, not auto-filed
Lead in Triage≥85% gate Disputeto Google Credit rateby market/hour Budget gatequality-weighted loop
Triage → dispute → credit rate → budget gate → back into triage. Quality and spend are one signal.
Mechanism 03

The system knows what it's allowed to do alone

Every action is typed by confidence × reversibility. The riskier and less reversible a move, the more human sign-off it needs. Automation is earned, not assumed.

Auto-execute

High confidence · reversible

Pace a budget within guardrails, reply to a new lead, request a review. Done instantly, logged, and scored later.

Approval required

Material · slower to undo

Large budget swings or structural changes are queued for your LSA manager to approve before they go live.

Alert only

Low confidence · high stakes

Ambiguous leads and "two probes came back dark" situations stop and page a human — because sometimes the problem isn't budget.

Proof, not promises

Automation that earns its autonomy

CallRadius keeps an outcomes ledger: every budget change it makes is scored about a week later — win, loss, or neutral — against real leads and booked revenue.

When a rule starts losing more than it wins, the system automatically demotes it to advisory until its record recovers. The AI has to keep proving each decision helps to keep making it.

The LSA tool that grades its own decisions
"Demand surge → +20% budget" — scored 8 days after each change W1W2W3W4W5W6W7
Losses now exceed wins → rule auto-demoted to advisory
Budget sweet-spot discovery

Finds the exact budget where you start showing

On Local Service Ads, budget buys eligibility — there's a competitive threshold where your ads appear, and drifting below it can make them vanish. The engine actively probes up and down to discover that line, waits 48 hours to confirm each move on real data, and locks your budget just above it.

If two probes come back "dark," it stops and alerts a human — because sometimes the problem isn't budget, it's responsiveness or reviews.

minimum-showing budget locked
Probe down to the threshold, confirm on real data, lock just above it.
The intellectual property
One system. Three mechanisms
no competitor combines.

CallRadius is protected by a U.S. patent application (No. 64/063,539) covering a hierarchical rule-suppression budget engine with an AI-driven lead-quality feedback loop. The status is patent pending.

Hierarchical suppression

Protective budget rules automatically override growth rules through a boolean gate — solving the conflicting-signals problem that makes other engines oscillate.

AI credit-recovery loop

Confidence-gated lead triage submits feedback to Google, and the resulting credit rate feeds back into the budget quality gate.

Risk-tiered autonomy

Every action is typed by confidence × reversibility — auto-execute, approval-required, or alert-only — so the system knows what it's allowed to do alone.

See it running

Watch the closed loop work on a live account

Open the interactive demo, or run the free LSA score to see where your account stands today.