PULSE AI-Powered Media Insights

Stop scouring dashboards.
Start reading insights.

Pulse is an AI agent that scours your media data for issues so you don't have to. It reads every metric, compares every run, and hands you the story — not the spreadsheet.

Enable Pulse
3-stage
Guardrailed pipeline
▲ No hallucinations
Run-over-run
Long-term memory
▲ Gets smarter
$0
Per page view
▲ Batch-powered
Powered by Enterprise-grade AI · Long-term memory · Self-service
AI Insights
BETA
Active Updated 2 min ago
Error count increased to 79 fires (+6.9% vs previous run). The two dominant failure modes are provider-side timeouts distributed across Chrome, Edge, and Firefox. Google Ads conversion tag showing intermittent 503s since 14:00 UTC.
Failure Score
0.21%
vs 0.18% prev
Key RTD
purchase
conversion tag
Findings
3
action items
38,498
Total Tag Fires
31,666
Successful Fires
6,955
Privacy Blocked
Why Pulse

Stop being reactive. Start being proactive.

Manual dashboards find issues after they happen. Pulse finds them while they're forming — and tells you exactly what to do.

Traditional Dashboard
Reactive. Manual. Starts from scratch every time.
You build dashboards to find issues. Then you spend hours figuring out what the numbers mean.
You open the dashboard. You see numbers.
Charts show state — not what changed, why it changed, or whether you should care.
You spot something off. You dig manually.
Export to sheets, cross-reference tags, check logs — hours of tribal knowledge applied per incident.
Next time you open it, you start from scratch.
No memory of what you checked last time. No context carried forward. Every session is day zero.
Alerts fire when thresholds break.
You find out after the damage is done — conversion data already leaked into ad platforms with bad values.
Analysis lives in your head.
When you're out, the team has no institutional memory of what "normal" looks like for this dashboard.
Pulse · AI-Powered Insights
Proactive. Automated. Remembers everything.
An AI agent scours your data on a schedule, surfaces the story, and builds on what it learned last time.
Dashboard tells you what changed — and why.
Pulse writes the analysis. Error count up 6.9%? It explains the two dominant failure modes, the affected browsers, and the timeline.
Run-over-run memory. It learns.
Every execution remembers what it saw last time. Facts accumulate. Insights evolve. Day-over-day trends are detected automatically, not manually.
Three-stage guardrailed pipeline.
Investigator queries the data. Reviewer blocks hallucinations and speculative claims. Layout renders clean, structured blocks. Not a chatbot — a controlled system.
Proactive — before thresholds break.
Pulse catches drift and anomalies at the pattern level, not the threshold level. Issues surface when they're trending, not after they've shipped.
Institutional memory, not tribal knowledge.
When you're out, the AI's memory is still there. Facts, trends, and past observations persist across runs — your team never starts from zero.
VS
Core Capabilities

Reduce hours of scouring to seconds of reading

Pulse does the work you'd spend a morning on — querying data, comparing metrics, spotting anomalies — and hands you informed story points instead of raw numbers.

Long-term memory

Every run, Pulse reads its memory — immutable facts with 30-day TTL, evolving insights, and a self-directed focus that tells it what to investigate next. It compares what it sees now against what it saw last time and calls out what changed.

Fact + insight evolution
R1
8
Facts stored
Baseline
R2
12
Facts (deduped)
+4 new · 1 resolved
R3
11
Facts (evolved)
Trend: 15%→18% ↑
Proactive intelligence

Runs on a recurring batch schedule — not on page load. Your dashboard has fresh insights waiting before you even open it. No per-view AI cost.

Every N hours
Day-over-day trending

Each execution compares against previous runs. Purchase fires dropped 15% → 18%? Pulse catches the trend and calls it out in context.

Run-over-run
No hallucinations

Three-stage pipeline: Investigator queries data, Reviewer blocks speculative claims, Layout renders structured blocks. Not a chatbot.

3-stage · guardrailed
Self-service

Enable Pulse per project from settings. A batch job is created from a cluster template — no tickets, no DevOps, no waiting.

One toggle
Your dashboard

Insight blocks inject above your existing widgets. Metric cards, analysis text, health checks — right where you already look.

Zero context switch
How Pulse Works

From schedule to insight in seconds

01
Schedule
Worker · Cron batch

A recurring batch job fires on your chosen cadence — every 4, 8, or 12 hours. The worker picks up the job with full vendor and project context, reads memory from the previous run.

Every N hrs
02
Investigate
Trino · Data queries

The Investigator runs the dashboard-insights runbook — querying for tag fire health, conversion rates, failure patterns, and anomalies. 6+ tool calls, fully data-grounded.

~26 sec
03
Review
Enterprise AI · Guardrails

The Reviewer stage blocks hallucinations, speculative claims, legal language, and invented deadlines. Inverted guardrails define what IS allowed. Only data-grounded observations survive.

~2 sec
04
Render & remember
Storage · Blocks + memory

Layout renders structured insight blocks for the dashboard. The memory extractor updates ai_memory with new facts, evolved insights, and a self-directed focus for the next run.

~2 sec
End-to-end: ~30 seconds
Zero manual analysis required
Memory persists across every run
Coverage

What Pulse watches for you

Cycle 1 covers the two areas that cost you the most when they break — revenue attribution and tag configuration health.

Revenue & Attribution Risk
Catches the failures that silently bleed ad spend and misattribute conversions.
Broken conversion tags
Detects tags that stopped firing or are returning provider-side errors (503s, timeouts).
Zero-fire detection
Flags purchase, add-to-cart, or lead tags that recorded zero fires in the analysis window.
Dedup anomaly check
Spots duplicate conversion events across tag providers that inflate reported ROAS.
Fire rate trending
Compares purchase fire volume run-over-run and flags drops above a configurable threshold.
Configuration Health
Surfaces tag config problems that silently rot your data collection over time.
Expired tags still live
Tags past their expiry date that are still actively firing — consuming budget and polluting data.
Unpublished changes
Draft changes sitting in the container that were never published — intent without execution.
Zombie tags
Tags configured and enabled but recording zero fires over the window. Active config, no output.
Orphaned providers
Tag providers with no active tags connected — leftover integrations from past campaigns.
Coming in future cycles: consent intelligence · traffic trends · event funnel quality
Real Output

What Pulse actually tells you

These are the kinds of insight blocks that show up on your dashboard — written by AI, grounded in your data, and compared against previous runs.

Revenue risk

Purchase tag fire rate dropped to 142 fires — down from 168 in the previous run. Google Ads conversion tag returning intermittent 503s since 14:00 UTC across Chrome and Edge.

Run #7 · 2 hrs ago ↓ 15.5% vs prev run
Config health

3 zombie tags detected — configured and enabled in GTM but recording zero fires over the last 72 hours. All three are Meta CAPI tags linked to inactive campaigns.

Run #7 · 2 hrs ago Unchanged from run #6
Trend · memory

Privacy-blocked fire rate has been climbing steadily: 16.2% → 17.1% → 18.0% over the last three runs. Correlates with increased Safari traffic from the iOS campaign launched Monday.

Run #7 · 3-run trend ↑ 1.8% over 3 runs
Resolved

Dedup anomaly on LinkedIn Insight Tag cleared. Duplicate conversion events dropped from 23 to 0 after the container republish at 09:15 UTC. No further action needed.

Run #7 · 2 hrs ago ↓ 23 → 0 duplicates

Stop reacting. Start knowing.

Enable Pulse and trade hours of manual dashboard scouring for seconds of informed reading. Proactive, memory-powered, on the dashboard you already open.