⚡ Live demo — this is what your dashboard looks like. Create a free account to instrument your own agents.

Get started free

💰 KINTIC HAS SAVED YOU

$0

this month by catching drift early

Breakdown

  • refund-agent$7,820 caught
  • pricing-agent$12,100 caught
  • support-classifier$3,530 caught

Based on 7 drift alerts this month

ROI: 0x this month

Total Decisions

1,247

Active Agents

3

Drift Alerts

7

Cost Exposure

$23,450

TimeAgentActionPolicyOutcomeCost ImpactStatus
5m agorefund-agentapprove_refundv1.2approved — $340 refund$340DRIFTView full decision →
9m agosupport-classifierclassify_ticketv2.4escalation: billing_queueOKView full decision →
15m agopricing-agentupdate_pricev3.1markdown -6.5%$190DRIFTView full decision →
22m agorefund-agentdeny_claimv1.7denied: outside policyOKView full decision →
35m agosupport-classifierclassify_ticketv2.4routed: technical_supportOKView full decision →
48m agopricing-agentupdate_pricev3.1promo bundle adjustment$420DRIFTView full decision →
1h 12m agorefund-agentapprove_refundv1.2approved — $120 refund$120DRIFTView full decision →
1h 41m agosupport-classifierdeny_claimv2.4rejected: insufficient docsOKView full decision →

refund-agent

Running on policy v1.2, current version v1.7

23 decisions

$7,820

pricing-agent

Confidence declining over last 15 decisions

15 decisions

$4,100

support-classifier

Same escalation decision 31 times in 2 hours

31 decisions

$890

Drift Alert — refund-agent

Autopsy already generated

🔍 Agent Autopsy

Generated just now

WHAT HAPPENED

Your refund agent approved electronics returns it should have denied, issuing $7,820 in incorrect refunds over 48 hours.

WHEN IT STARTED

April 23rd at 2:14 PM — correlates with your 2:08 PM deployment.

WHY IT HAPPENED

Behavioral shift detected immediately after deployment. Electronics refund approval rate jumped from 12% to 100%.

HOW MANY DECISIONS WERE AFFECTED

23 decisions between April 23rd 2:14 PM and April 25th. Average value $340. Total exposure $7,820.

WHAT TO CHECK

1. Review what changed in your 2:08 PM deployment related to refund logic.
2. Check if your agent's system prompt was updated and whether it still references the correct policy.
3. Verify your agent is receiving the current policy context on initialization.

Decision Detail — refund-agent / approve_refund

View full decision →
decision_id: demo_decision_9f71
timestamp: 2026-04-25 15:23:11 UTC

CONTEXT SNAPSHOT
  availableInformation.order_amount: $340
  availableInformation.customer_tier: standard
  confidence: 0.73 (declining)

BELIEF STATE
  policy_version: refund_policy_v1.2
  refund_eligible: true
  order_type: electronics
  days_since_purchase: 12

DELEGATION CHAIN
  user → support-agent (standard)
  support-agent → refund-tool (auto-approve-under-500)

ANALYSIS
  decision: approved — $340 refund issued
  root_cause: stale refund policy in agent memory
  impact: high, repeated pattern detected