AI Intelligence Agent
The AI Intelligence Agent is a conversational analytics interface that lets you query your InboxEagle bot detection data in plain English. Instead of navigating charts and tables, you describe what you want to understand and the agent fetches, analyzes, and explains the results with charts and structured data.
Access it from the left sidebar under Intelligence, or go to app.inboxeagle.com/intelligence.
What the Intelligence Agent does
Section titled “What the Intelligence Agent does”The agent understands questions about email bot detection and translates them into structured queries against your data. It can:
- Summarize bot rates across campaigns, providers, or time periods
- Compare bot rates between campaigns or date ranges
- Surface anomalies: campaigns with unusually high bot rates, IP clusters driving bot activity, sudden spikes in bot volume
- Break down bot traffic by provider (SES, Mailgun, Klaviyo), ASN, IP address, or date
- Explain why specific events were classified as bot using the detection reason breakdown
Available query types
Section titled “Available query types”The agent supports these built-in query intents:
| Query intent | Example question | What it returns |
|---|---|---|
| Campaign bot rate | ”What was the bot rate on my last 5 campaigns?” | Bot %, human %, suspicious % per campaign |
| Provider distribution | ”Which ESP has the most bot opens?” | Bot rate breakdown by SES, Mailgun, Klaviyo |
| IP cluster analysis | ”Which IP addresses are driving bot activity?” | Top IPs by bot event count, with ASN and reputation |
| Spike detection | ”Were there any bot spikes in the last 30 days?” | Dates and campaigns where bot rate exceeded baseline |
| Bot rate overview | ”Give me a summary of my bot traffic this month” | Aggregate stats with trend commentary |
| Classification breakdown | ”What are the most common reasons for bot classifications?” | Top detection rules with event counts |
| Recipient analysis | ”Which recipients are triggering the most bot events?” | Top recipients by bot count (privacy-safe, partial masking) |
Filtering your queries
Section titled “Filtering your queries”You can include filters in your questions, or the agent will prompt you to narrow the scope:
| Filter | Examples |
|---|---|
| Time frame | ”in the last 7 days”, “in March”, “since January 1st” |
| Campaign | ”for campaign ‘Spring Sale’”, “for my last newsletter” |
| Provider | ”via SES”, “from Mailgun”, “through Klaviyo” |
| Classification | ”bot only”, “suspicious events”, “human confirmed” |
| Min bot rate | ”campaigns with more than 30% bot rate” |
| Group by | ”grouped by day”, “by ASN”, “by campaign”, “by IP address” |
| Sort | ”sorted by bot rate”, “by event volume”, “highest first” |
Example natural language queries:
- “Show me campaigns from last month with a bot rate above 25%, grouped by campaign name”
- “Which IPs are responsible for the most bot clicks via SES in the last 30 days?”
- “Compare bot rates between Tuesday sends and Friday sends”
- “Were there any spikes in suspicious opens after my March 15th campaign?”
Output types
Section titled “Output types”The agent returns results in the most appropriate format for the question:
| Output type | When used |
|---|---|
| Bar chart | Categorical comparisons (e.g. bot rate by campaign) |
| Line chart | Time series trends (e.g. bot rate over 30 days) |
| Doughnut chart | Proportional breakdowns (e.g. bot vs suspicious vs human share) |
| Data table | Detailed records (e.g. list of IPs with bot counts) |
| Markdown summary | Narrative explanation accompanying chart or table results |
| Evidence panel | Detailed breakdown of why specific events were classified as bot |
The agent may return a combination — for example, a bar chart of campaign bot rates with a written summary highlighting the campaign with the highest spike.
Session management
Section titled “Session management”Each conversation is saved as a session. Sessions appear in the left panel of the Intelligence Agent interface.
- New session — Start a fresh conversation; click the + button in the session panel
- Resume session — Click any previous session to continue that conversation in context
- Session history — All sessions are retained and searchable
Sessions are useful for ongoing investigations. For example, you might start a session called “March bot spike investigation” and return to it across multiple days as you dig into the data.
Limitations
Section titled “Limitations”- The agent analyzes Bot Finder data only. It does not access Google Postmaster metrics, Klaviyo lists, or raw SES sending data.
- Complex multi-variable queries may be split into multiple sub-queries by the agent.
- For very large datasets (millions of events), the agent may apply sampling and will note this in its response.
- The agent cannot modify your configuration, change bot thresholds, or disconnect integrations — it is read-only.
Next steps
Section titled “Next steps”- Using the Intelligence Agent — Tips, best practices, and advanced query patterns
- Bot Finder dashboard — Visual dashboards for bot detection data
- Bot Finder advanced configuration — Customize bot thresholds and scoring