Invoices that file themselves
- 1Bills logged in accounting with the PDF attached
- 2Due-date reminders before anything is late
- 3Bank-detail changes flagged as high risk
Anthem / Email BrainLaunching Q1 2027
Screens every message for scams, sorts it, runs the right workflow in your tools and drafts the reply. You stay in control.
Inboxhello@your-company.example
agent onInvoices0Support0Leads0Blocked0
Your team did not join to copy invoice numbers, chase order statuses and delete phishing. Email Brain does the reading, filing and first drafts, so people spend their time on the replies that need a human.
How a message moves.
Every email follows the same accountable path. Switch between a legitimate email and a phishing attempt, or click a step to inspect it.
Drop in an email. Watch it work.
An interactive demo with sample messages. Pick one and follow each stage, from the scam check to the drafted reply.
Pick a sample email
ReadScam checkCategoriseExtractMCP toolsReply
From Northwind Supplies <billing@northwind-supplies.example>
Invoice INV-2291 for October
Hello, please find attached invoice INV-2291 for the October delivery of packaging materials. Total due: €4,860.00 by 15 November. Bank details are on page two. Kind regards, Eleni
Everything an assistant would do. None of the copy and paste.
Scam shield
Phishing stopped before anyone clicks.
Every message is checked first. Lookalike domains, spoofed senders, hidden links and sudden bank-detail changes raise the risk score and send the email to quarantine.
Categories in your words
Describe a category in a sentence. No rules engine, no regular expressions.
Workflows in plain language
Write what should happen. The agent turns it into durable steps that retry safely.
Bring your tools
CRM, helpdesk, calendar, accounting and chat, connected over MCP.
An agent per mailbox
Always on, with its own memory of threads, contacts and preferences.
Drafts in your tone
Grounded in the thread and your past replies.
Autonomy you dial in.
Start by watching what the agent would do. Move to drafts when you like what you see. Switch on auto-send only for the categories you trust.
The agent suggests, you act
Every message is screened and labelled, and suggested actions appear next to it. Nothing changes in your tools until you click.
- Scam screening on
- Workflow actions suggested
- Replies off
Work is done, replies wait for you
Workflows run in your tools and every reply is drafted and parked for your approval. Most teams start here.
- Scam screening on
- Workflow actions automatic
- Replies drafted
Trusted categories run end to end
Pick categories, like meeting changes or order updates, that send on their own. Everything else still waits for you.
- Scam screening on
- Workflow actions automatic
- Replies sent for trusted categories
Built for the teams that live in email.
Customers answered while they still care
- 1Tickets opened with order data and photos
- 2Replacement or refund prepared for approval
- 3Replies drafted in your support tone
Leads qualified before the coffee cools
- 1Deals and contacts created in your CRM
- 2Call slots proposed from your calendar
- 3The right owner notified in chat
Suppliers and schedules in sync
- 1Delivery updates matched to purchase orders
- 2Meeting changes rescheduled automatically
- 3Weekly digest of what the agent handled
Only the email that needs you
- 1A short morning brief instead of 200 unread
- 2Investor and partner threads always surfaced
- 3Everything else handled, logged and searchable
Secure by architecture, not by promise.
Email is where your business keeps its secrets. Isolation, least privilege and a complete audit trail are built into the platform.
Isolation per mailbox
Each mailbox is processed in its own runtime. Data never mixes between customers or mailboxes.
Least-privilege access
Scoped OAuth grants for mail and every connected tool, revocable at any moment.
Encrypted everywhere
TLS in transit, encrypted object storage for attachments and message data at rest.
Complete audit trail
What the agent read, decided, changed and sent, exportable whenever you need it.
Humans in the loop
Drafts by default, approval gates for sensitive actions, one-click undo where possible.
Your data, your models
We do not use your email to train models. Retention periods are yours to set.
Questions, answered.
The short version. Anything else, just ask.
When is Email Brain available?
We launch in Q1 2027. Waitlist members are first in line when we open access, and help us decide which integrations ship first.
Which mailboxes are supported?
Google Workspace, Microsoft 365 and standard IMAP mailboxes, including shared inboxes such as info@ and support@. You connect them with scoped permissions that you can revoke at any time.
Will it send emails without me?
Only if you want it to. Every reply starts as a draft for your approval. Once you trust the results you can switch on automatic sending for specific categories, and switch it off again whenever you like.
How does the scam shield work?
Each message is scored before anything else runs, using signals such as lookalike domains, spoofed senders, mismatched links, urgency language and unusual payment requests. High-risk messages are quarantined with their links disabled.
Which tools can it connect to?
Any tool that offers a Model Context Protocol server, which already covers many CRMs, helpdesks, calendars, accounting packages, file storage and chat tools. We can also build connectors for in-house systems.
Where is my data processed and stored?
Each mailbox is processed in its own isolated runtime, attachments live in encrypted object storage, and every automated action is written to an audit log you can export. The service is designed around GDPR from day one.
Can it write in our tone of voice?
Yes. Drafts are grounded in your past replies and a short style guide that you control, so they sound like your team.
Get your inbox back in Q1 2027.
Join the waitlist and be first in line. Tell us which mailboxes and tools you use, and we will prioritise the integrations that matter to you.