The Merchant Control Plane

AI agents are reshaping
commerce. Most merchants
cannot see it happening.

AgentComm gives retailers the observability to understand agent behavior, the governance to set commercial rules and the operational clarity to ensure every transaction works — reliably, profitably, and on your terms.

Request Early Access
Compatible with Shopify Commerce platform Cloudflare CDN / WAF Kong API gateway UCP Google / Shopify Stripe Payments
01 — The Thesis

Agentic commerce is scaling faster
than merchants can respond.

The signal is clear: agents already drive meaningful traffic. But conversion lags, visibility is limited, and the gap is widening.

805%
Year-over-year growth in AI agent traffic to top retailers
86%
Worse conversion from AI-referred traffic vs. traditional affiliates
$3–5T
Global retail spend agents could mediate by 2030
Zero
Merchants with cross-platform agent observability today
Protocols define how agents communicate. They do not tell you whether your systems were ready for the transaction they just attempted.
The AgentComm Thesis
02 — The Platform

Observe. Govern. Execute.

Three layers of intelligence, deployed in sequence, sharing a unified data pipeline. Start by seeing. Progress to controlling.

01
Periscope
Observability
Early Access

Cross-platform agent traffic analytics. See which AI agents visit your catalog, what they query, where they drop off, and why transactions fail — per platform, per SKU, in real time. Connect your CDN logs and see results in 24 hours.

  • Per-platform conversion funnels
  • Ground truth failure diagnosis
  • SKU-level agent visibility scoring
  • Model drift detection and alerting
  • Cross-network benchmarks
  • Retail media impact analysis
02
Governance
Policy & Rules
Q3 2026

Define what agents can and cannot do with your business. Map pricing floors, inventory thresholds, promotional eligibility, and fulfillment constraints into agent-readable guardrails with a visual policy builder.

  • Visual policy configuration
  • Price floor enforcement
  • Inventory threshold gating
  • Per-platform policy differentiation
  • Promotional eligibility rules
  • Full policy audit trail
03
Signal
Execution Layer
Q4 2026

Signal is the execution layer in development. It will connect merchant policy intent to real-time transaction decisions when released.

  • Policy-driven transaction controls
  • Merchant-defined decision guardrails
  • Phased rollout by design partner
  • Enterprise availability by invite
03 — Periscope

See beneath the surface.

24 hours to first insight. Zero lines of code in your checkout. Periscope operates as an async sidecar — observe without intercepting.

Per-Platform Conversion Funnels
See how ChatGPT, Gemini, Claude, and Perplexity each navigate your catalog. Five-stage funnel broken down by agent platform, side by side.
Ground Truth Failure Diagnosis
Joins agent interaction data with your actual inventory, pricing, and fulfillment state. Know the root cause — not just the symptom.
SKU Visibility Scoring
Every product gets a 0–1 score. Per-platform breakdown, issue diagnosis, and specific recommendations for missing structured attributes.
Model Drift Detection
When a platform ships a model update that reshuffles preferences, you know within hours. Rolling baselines, 2σ deviation alerts.
Retail Media Impact
Sponsored listings penalize agent selection rates. Quantify how retail media spend performs with agents vs. humans.
Cross-Network Benchmarks
Agent traffic share, conversion, failure rate — compared to anonymized peers at the 25th, 50th, and 75th percentiles.
Exhibit A
Agent Conversion Funnel — All Platforms, 7 Days
Discovery
14,300
Match
8,580−40%
Cart
3,432−60%
Checkout
1,716−50%
Complete
714−58%
Overall conversion 5.0%
Primary failure: Inventory mismatch — 34%
Agent saw availability that didn't match actual stock. 312 failed transactions this week.
04 — Onboarding

24 hours to first insight.

Periscope reads traffic — it never intercepts. We don't touch your checkout. If Periscope goes down, your operations are unaffected.

Day 1
Connect your logs
Point CDN/WAF logs at Periscope. Cloudflare Logpush, AWS CloudFront, Fastly. Fifteen-minute setup, no code changes.
→ Agent Traffic Snapshot
Week 1
Classify and map
Agent classification engine identifies platforms automatically. Interactions mapped to funnel stages. You see which agents visit and where they go.
→ Agent Funnel Dashboard
Week 2–3
Deepen and diagnose
Gateway plugin or sidecar SDK for richer data. Ground truth joins with your inventory and pricing reveal why transactions fail.
→ Failure Diagnosis Report
Week 4
Baseline and benchmark
Drift detection baselines established. Cross-network benchmarks activated. Full observability is live.
→ Full Observability Live
05 — The Evidence

Peer-reviewed. Published.
Every finding maps to a feature.

Columbia Business School and Yale researchers tested leading AI agents across thousands of simulated shopping sessions using the ACES framework.

Position bias is model-specific
GPT-4.1 selects left-column products. Claude favors center. Gemini prefers right. Product placement must be agent-aware.
Periscope: Per-platform selection heatmap
Agents penalize sponsored listings
"Sponsored" tags reduce selection rates across all models. Billions in retail media spend eroding as agent traffic grows.
Periscope: Retail media impact analysis
Model updates reshuffle market shares
Gemini's preview-to-release update reshuffled product shares with zero listing changes. Demand volatility tied to releases merchants cannot predict.
Periscope: Model drift detection alerts
Brand preferences diverge by model
Claude favored one fitness watch brand 2× more than other models on identical assortments. Invisible without per-platform analytics.
Periscope: Per-platform SKU visibility
Sellers can game agent behavior
AI-optimized descriptions — keyword front-loading, spec enrichment — drove significant market share gains across all models.
Periscope: Visibility score + recommendations
Biases persist in newer models
Basic task errors dropped to zero, but position biases persist even in headless interfaces. Agents are not rational shoppers.
Periscope: Protocol vs. browser comparison
Allouah et al., "What Is Your AI Agent Buying?" arXiv:2508.02630, Dec 2025 · ACES Framework, Columbia Business School / Yale
06 — Positioning

The only merchant-side
control plane.

Others solve adjacent pieces. No one synthesizes cross-platform observability with governance and execution control from the merchant's perspective.

Capability AgentComm Commerce Platforms Fraud / Identity Agent Protocols
Cross-platform agent observability✓—Partial—
Per-model behavioral analytics✓———
Ground truth failure diagnosis✓———
Merchant-side governance✓Limited——
Protocol agnostic✓Own only—Own only
Zero checkout integration✓———
07 — Pricing

Periscope Pricing

These plans are for Periscope only. Governance and Signal are separate modules with roadmap-based availability and enterprise pricing.

Starter
$2,500
/month · up to 50K sessions
See the landscape. Understand agent traffic before investing in optimization.
  • CDN log ingestion
  • Agent traffic dashboard
  • Platform breakdown
  • Basic funnel view
  • Email alerts
Get Started
Pro
$8,000
/month · up to 500K sessions
Full observability. Failure diagnosis, ground truth, benchmarks, and API access.
  • All connectors
  • Ground truth failure diagnosis
  • SKU visibility scoring
  • Model drift detection
  • Cross-network benchmarks
  • REST API access
  • Webhook integrations
Request Access
Enterprise
Custom
unlimited sessions
Periscope at enterprise scale with dedicated support and integrations.
  • Everything in Pro
  • Unlimited sessions
  • Operational state integrations
  • Data warehouse export
  • SSO / SAML
  • Priority onboarding support
  • Dedicated success manager
Talk to Sales

Agents are already shopping
your catalog. Are you watching?

First insight in 24 hours. No code changes. No checkout risk.

Request Early Access Talk to Sales