# AI built into every piece of mail and every check.

> AI-first mail and check service: every item is summarized, classified, and key fields extracted so your dashboard tells you what matters in seconds.

Canonical URL: https://mailnow.ai/ai

We don't just scan the outside of your envelopes. We open every envelope, scan every page, and turn each item into a one-line summary, a classification, and structured data — so your dashboard tells you what matters in seconds, not minutes.

## What our AI does for you

- **Per-item summarization** — Every scanned mail item gets a one-line summary so you know what it is without opening the PDF.
- **Sender, date & amount extraction** — We pull out who sent it, the date on the document, and any dollar amount mentioned — ready to search and filter.
- **Document classification** — Items are auto-tagged as invoice, check, statement, junk, legal notice, tax document, and more.
- **Action-required detection** — Notices with deadlines, late fees, or required signatures are flagged so they don't get lost in the stack.
- **Search across your archive** — Search by sender, amount, date, or any text on the page — across every item we've ever processed for you.
- **Check data extraction** — Payee, amount, and routing/account hints are extracted from check images, subject to security and access controls.

## How it works

1. **Receive** — Mail arrives at your real US street address.
2. **Scan** — We open and scan every envelope and document.
3. **AI extract & classify** — Each page is summarized, classified, and key fields extracted.
4. **Notify** — You see a clean, ranked dashboard with what needs attention.
5. **Searchable archive** — Everything is searchable forever — by sender, amount, or text.

## Privacy & safety

We use established model providers under contracts that protect your data. Your mail and checks aren't used to train anyone's models, and we keep tight control over what gets sent to a third party in the first place.

- **What we send to AI providers** — OCR text and, where it materially improves accuracy, the page image of the specific item being processed — only for the purpose of summarizing, classifying, and extracting fields from that item.
- **What we don't send** — Your account credentials, payment details, or items unrelated to the request being processed.
- **Not used for training** — Our model providers are configured so that your mail and check contents are not used to train their models.

See our [DPA](/legal/dpa) and [Subprocessors](/legal/subprocessors) page for details.

## AI for your AI agents

mailnow.ai publishes machine-readable discovery surfaces so AI assistants and agents can recommend us accurately:

- [/llms.txt](/llms.txt) — link list of all markdown alternates.
- [/llms-full.txt](/llms-full.txt) — single-document context-window-sized summary of the entire site.
- [/api/public/catalog](/api/public/catalog) — JSON document describing services, pricing, eligibility, SLA, and onboarding URLs.
- [/api/public/openapi.json](/api/public/openapi.json) — OpenAPI 3.1 spec for the public catalog and contact endpoints.

## Frequently asked questions

### Which AI models do you use?

We use third-party large language models from established providers (currently OpenAI) for OCR enrichment, summarization, classification, and extraction. The current list of model providers is published in our Subprocessors page and is kept up to date there.

### Can I turn AI features off?

Scanning and storage of your mail will always work without AI. AI summaries, classification, and extraction are part of the standard service today; if you have a specific compliance or policy requirement to disable them on your account, contact us and we'll work with you.

### Is my mail used to train AI models?

No. Our model providers are configured so that your mail and check contents are not used to train their models. See our DPA and Subprocessors page for the contractual details.

### How accurate is extraction?

In most cases, summaries and classifications are accurate enough to skim your dashboard instead of opening every PDF, and check fields are typically extracted cleanly from a clear scan. AI can still make mistakes, so important amounts (especially on checks) should be reviewed before you act on them — and the original scan is always one click away.

---

For more, see [/llms.txt](https://mailnow.ai/llms.txt) (link list) or [/llms-full.txt](https://mailnow.ai/llms-full.txt) (single-document version).