Last updated: March 29, 2026
InflowMail uses AI to analyze every incoming email — classifying, prioritizing, and surfacing insights so you can focus on what matters.
When a new email arrives, InflowMail processes it through a two-tier classification pipeline designed to be both fast and accurate:
Rules + Database Intelligence (Free, Instant)
Most emails (~90%) are classified instantly using built-in rules, sender reputation data, and domain intelligence already stored in your account. This step is free and requires no AI API calls.
AI Classification (Paid Plans)
When the rule engine's confidence is below 85%, the email is sent to an AI model for deeper analysis. The AI result is cached back into domain reputation data so future emails from the same sender are classified for free.
User Rules Applied Last
After AI classification, any user-created rules are evaluated and applied. Your rules always override what the AI decided.
Self-improving system: Every time the AI classifies a new sender, the result is cached into domain reputation data. Over time, more emails are handled by the free tier, reducing AI costs while maintaining accuracy.
Every email receives a Priority Tier — High, Medium, or Low — so you always know what needs your attention first.
| Tier | What It Means | Examples |
|---|---|---|
| High | Needs your attention soon. Action required or from an important sender. | Client requests, urgent deadlines, VIP contacts |
| Medium | Worth reading but not urgent. Informational or routine correspondence. | Team updates, order confirmations, calendar invites |
| Low | Can wait. Background information, bulk mail, or purely informational. | Newsletters, promotional offers, social notifications |
Priority is determined by combining multiple signals:
Manual override: You can always change an email's priority tier manually. Once overridden, the AI will not change it back. Use the "Why?" link to understand why the AI chose a particular tier.
Every email is automatically assigned a category so your inbox stays organized without manual effort.
Primary
Direct messages from real people
Updates
Order confirmations, receipts, shipping
Social
Social network notifications
Promotions
Marketing emails and deals
Newsletters
Subscriptions and digests
Forums
Discussion groups and mailing lists
Categories are visible in your inbox tab bar (All | Focus | Other | Bills | Receipts | Newsletters) and can be used as conditions in email rules.
For emails processed by the AI tier, InflowMail generates two additional pieces of information:
AI Summary
A concise, one-line summary of the email's content. This appears in your inbox list view so you can quickly understand what the email is about without opening it. Especially useful for long threads and detailed messages.
Suggested Action
When the AI detects that an email requires action, it flags it as actionable and suggests what you might want to do — such as "Reply," "Review attachment," or "Schedule meeting." These suggestions appear alongside the email to help you process your inbox faster.
Paid plans only: AI summaries and suggested actions require a plan with AI features enabled. On the free tier, emails are still classified and prioritized using the rules + database intelligence engine.
InflowMail automatically scans every email for signs of phishing and assigns a phishing score (0 to 1.0). This analysis runs independently of AI classification and uses heuristic detection techniques.
What the phishing detector checks:
Display name mismatches — when the sender's display name does not match their actual email address (e.g., "PayPal" sending from a non-PayPal domain).
Suspicious links — URLs that point to known phishing domains, IP addresses, or use URL shorteners to hide their destination.
Lookalike domains — domains designed to impersonate legitimate services (e.g., "paypa1.com" instead of "paypal.com").
Urgency language — phrases like "act now," "account suspended," or "verify immediately" that are commonly used in social engineering attacks.
Suspicious attachments — risky file types (e.g., .exe, .scr, .bat) that are commonly used to deliver malware.
| Score Range | Risk Level | What You See |
|---|---|---|
| 0.0 – 0.3 | Low | No indicators shown. Email appears safe. |
| 0.3 – 0.6 | Medium | Caution banner with specific indicators. |
| 0.6 – 0.8 | High | Warning banner with details about suspicious elements. |
| 0.8 – 1.0 | Critical | Strong warning. Likely a phishing attempt. |
Each day, InflowMail generates a personalized briefing that summarizes your inbox activity. Think of it as a morning report for your email.
Your daily briefing includes:
Priority items — the most important emails that need your attention, ranked by priority and action type.
Awaiting reply — sent emails where you are still waiting for a response, with how many days you have been waiting.
Catch-up items — older emails you may have missed that still need attention.
Stats and insights — new email count, category breakdown, top senders, and trends across your accounts.
You can browse past briefings in your briefing history. Briefings are generated on demand when you open the briefing view.
InflowMail supports multiple AI providers. You can use the default InflowMail AI, choose a specific provider, or bring your own API key for full control over costs and model selection.
| Provider | Supported |
|---|---|
| Anthropic (Claude) | Default provider |
| OpenAI (GPT) | Supported |
| Google (Gemini) | Supported |
| xAI (Grok) | Supported |
For detailed setup instructions, API key configuration, and provider comparison, see the Bring Your Own AI documentation.
Your email rules always take priority over AI classification. The processing order is deliberate:
AI classifies the email (category, priority, summary)
Your rules are evaluated against the email
Rule actions override AI decisions (priority, category, labels, etc.)
This means you can fine-tune the AI's output without disabling it. The "Why?" trace panel in your inbox shows both the AI's original decision and any rule overrides so you always understand how an email was classified. Learn more in the Email Rules documentation.
InflowMail is designed to minimize the data shared with AI providers. When the AI tier is used for classification:
Metadata only. The AI processes the email subject, sender name, sender address, and a short content snippet — not the full email body.
No training. Your email data is not used to train AI models. API requests are processed and discarded by the provider.
Encrypted storage. OAuth tokens and any API keys you provide are encrypted with AES-256 at rest.
Your choice of provider. With Bring Your Own AI, you control which provider processes your data and can use your own API key.
For our full privacy practices, see the Privacy Policy and Security documentation.
The two-tier system achieves high accuracy by combining rule-based intelligence (which learns from AI results over time) with AI analysis for uncertain cases. Classification improves the longer you use InflowMail, as domain reputation data grows. If the AI gets something wrong, you can create a rule to correct it for all future emails.
The free-tier rule engine always runs to provide basic classification and priority. If you do not want AI API calls for deeper analysis, you can stay on the free plan — the rule engine handles most emails without any AI provider calls. You can also create manual rules to override any classification you disagree with.
No. The AI receives only the email subject, sender information, and a short content snippet for classification. Full email bodies are stored in your database (or optional S3 storage) and are never sent to AI providers for classification purposes.
If an AI API call fails, InflowMail falls back to the rule-based classification result. Your emails are still categorized and prioritized — just with slightly lower confidence for unfamiliar senders. The system retries AI classification in the next enrichment batch.
Have questions about AI features?