See which keyword actually printed money.
Conversion intelligence per agent and per bot. Keyword Revenue P&L joins Google Ads spend (Microsoft coming) to the visitors who actually converted. ROAS by keyword, sortable, exportable.
Last 30 days
482
Conversations
23.4%
Conversion
$54.2K
Revenue
Visitors raising “pricing concerns” aren't converting.
Suggested directive: acknowledge the budget reality, offer a smaller scoped start or installment plan, show flexibility before re-anchoring.
Bot vs each agent · last 30d
Your AI gets sharper from every chat that closes (or doesn't).
ZenTalk reads every closed conversation, extracts the topics raised, the objections that came up, and the sentiment trajectory. Patterns you'd never spot manually surface as one click directives. Apply to bot, the next reply uses what you just learned.
AI bot AND human agent analysis
Don't just measure the bot. Every agent on the team gets attribution: who closes warm chats, who handles refund objections best, who's the team's top performer this month. Manager grade visibility, no separate analytics tool.
Apply to bot, in one click
A "loss pattern" card with a hand tuned prompt addition lands in the dashboard. Click Apply, it lands in Site.salesAgent.injectedInsights, next bot reply incorporates it naturally. No restart, no redeploy.
Real cost, real budget
Hourly Gemini Flash analyser at about $0.0001 per conversation, capped per workspace by GEMINI_DAILY_CAP. The daily card generator is pure SQL with no AI cost. The whole system runs at fractions of a dollar per workspace per month.
Evidence on every card
Each card links to the source conversations that triggered it. Every claim is auditable; every directive is grounded in real visitor language, not hallucinated trend lines.
This week
127
IPs blocked
127
Synced to Google
$432
Saved
198.51.100.42
threat intel: VPN (risk 88)
Google Ads · synced to 4 campaigns
198.51.100.42 added to Excluded IPs
~3.2s after block · auto
Stop paying Google for clicks you already know are junk.
ZenTalk classifies every visitor IP. Proxy, bouncing bot, known threat, manual block. Then pushes the result straight into your Google Ads campaigns' Excluded IPs list within seconds. The same junk traffic that was about to drain your daily budget never gets a second click charged.
Click fraud gateway in front of your landing page
Industrial threat intel feeds (10k+ IPs refreshed daily) plus VPN and Tor scoring. Bots, low quality proxies, and rapid bouncers get a 403 before they touch your site.
Auto block rules per site
Rapid bounce uses a configurable window plus threshold (default: 2 sessions in 2 seconds blocks the IP). Threat intel match is instant. All toggleable per site so each brand has its own risk appetite.
Connect Google Ads with one OAuth click
Per site OAuth to your Google Ads manager (MCC). Pick the customer accounts and campaigns that should mirror this site's blocklist. Encrypted refresh tokens. Only the IP exclusion scope is touched. No bid changes, no creative writes, no spend access.
Every block and unblock fans out within seconds
Manual block in your blocked IPs page? Synced. Auto block from rapid bounce? Synced. Unblock to give a friend a second chance? Removed from Google. The reconcile is idempotent. Your manually added Google Ads exclusions stay untouched.
Per IP click logs plus reconciliation
Every gateway click logs the verdict, source IP, ad network click ID (gclid, msclkid, and so on), and the brand it landed on. Marketers reconcile against Google Ads reports daily, so drift, double charges, and partner network leaks all surface.
Pay the team that closed the deals.
Attendance, monthly payroll, salary slips: same workspace, same people. See HR + payroll.