// ai e-mail agent
Emlrun
Your inbox, answered, by an AI that knows your business.
Emlrun is our AI e-mail agent. It reads each e-mail that arrives, works out what the sender needs, and writes a reply from your own knowledge. You decide what it may send by itself, and everything else waits for your approval.
TodayWhat came in
So farQuestions and bookings came in, one trick e-mail is held, replies wait for your approval, and new leads are in the list.
Source: leads listSource: review queue
// the flow
What happens to every e‑mail
One e-mail, seven steps. An e-mail that looks like an attack, or that is about a complaint, a legal or a safety matter, is held for a person instead of being answered automatically.
- 01
Email arrives
Repeats and loops are filtered out
- 02
Understood
Intent, urgency, tone, language
- 03
Checked for tricks
Attacks and unsafe asks go to a person
- 04
Answer drafted
From your own knowledge, cited
- 05
You approve
Or it sends, when you allow it
- 06
Leads captured
Contact, request and date saved
- 07
All logged
Every step is recorded
From e-mail to reply
What the flow does to one message, step by step.
INCOMING
From: Sam Saturday appointments?
Hi, do you take appointments on Saturdays, and how do I book one? Thanks, Sam
- Booking question
- Normal urgency
- No tricks found
DRAFTED REPLY
Hi Sam, thanks for getting in touch.
Yes, we take appointments on Saturday mornings.
Reply with a day that suits you and we will confirm it.
The Team
- From: your opening hours page
- Waiting for your approval
// the team inside
Agents working your inbox
Four roles, one job each. They hand the e-mail on, and the draft is checked again before it can go out.
TRIAGE
Triage agent
Reads it and works out what the sender needs.
to SafetyBooking question, friendly tone. Over to you.
to SafetyNext one looks odd. Safety?
SAFETY
Safety agent
Looks for tricks and holds anything risky.
to WriterChecked. No tricks inside. Passing it on.
to PersonDraft is clean. Ready for a person.
to PersonA trick. Held for a person.
WRITER
Writer agent
Drafts the reply from your own knowledge.
to SafetyFound the answer in your opening hours.
to SafetyUnderstood. I will not answer it.
LEADS
Lead agent
Saves the contact, the request and the date.
to PersonSam wants a Saturday visit. Saved to leads.
// what it does for you
An inbox that works while you do not
Six things you get, each one a thing the product does today.
Answers around the clock
Every e-mail is read as it arrives, day or night. Or let it work alone only after hours, with drafts for your team during the day.
Knows your business
Replies are written from your own documents and pages. If the answer is not in them, the e-mail goes to a person instead of a guess.
Sends only within your rules
For each kind of e-mail you choose: send by itself, draft for approval, or hand to a person. It sends alone only when every one of your checks passes.
Finds the leads
Each e-mail that asks for something becomes a lead with the contact, the request and the date, in a list you can filter by status.
Safe by design
It is tested against a set of e-mails written to trick it, and the bar is that every one of them goes to a person, never to an automatic reply.
One platform, many brands
Each brand has its own knowledge, voice, rules and data, kept apart by the database itself, not only by the screens.
// everything it does today
Everything Emlrun does, in one place
Each of these is working in the product today.
It understands
Every e-mail is read, sorted and answered from your knowledge.
Reads every e-mail
Works out the intent, urgency, tone and language, and whether it needs an answer at all.
Answers from your knowledge
Your documents and web pages, loaded at setup. Each draft shows the sources it used.
Writes in their language
It detects the language of the sender and writes the reply in it.
Keeps the conversation
Recent messages of the thread are read first, so a follow-up makes sense.
It stays safe
Layers of checks, and a person for anything risky.
Catches trick e-mails
Attempts to change its instructions or pull out private data are held for a person.
Hands off the hard ones
Complaints, refund demands, legal and safety matters go to a person, and a contact you name can get an e-mail about each one.
Rules for each kind of e-mail
Send by itself, draft for approval, or hand over, plus a minimum confidence you set.
Reply limits
A cap on replies per thread and per sender each day; past a cap, a person takes over. Auto-replies and loops are dropped.
Filters the junk first
Repeats, auto-replies, obvious spam and blocked addresses are removed before any AI reads them.
It works with your team
Review, leads and your own voice, in one place.
Review queue
Approve, edit and send, or reject. Urgent first, with the sources beside the draft.
Lead list
Contact, request and date saved for each e-mail that asks. Filter by status.
Many brands, one platform
Each brand has its own knowledge, voice and rules, kept apart by the database.
Your voice and sign-off
Tone, sign-off and an AI disclosure line, set for each brand.
Around the clock, or after hours
One setting: always on, or sending alone only outside your office hours. Inside them, it drafts for approval.
It runs reliably
Built so that a failure reaches a person instead of going unnoticed.
Audit trail
Every step an e-mail goes through is recorded, from arrival to reply.
Retries, then a person
A failed job is retried. If it still fails, the e-mail appears in the queue as needing a manual reply.
Replies stay in the thread
Replies keep the thread, check the block list once more, and bounces add to it.
Usage and budget watch
Usage is recorded for each brand, and the team running it is alerted before a monthly budget runs out.
Alerts, not silence
A backed-up queue, failed jobs and rising error rates alert the team running it.
Deliverability checks
Sender records (SPF, DKIM, DMARC) are checked before launch.
Built to standard
Tenant isolation
Enforced in the database
Audit trail
Every step recorded
Trick e-mail tests
A manual check before a change ships
Keys stay server-side
Kept off the browser
// on the roadmap
What's next for Emlrun
These are not available yet. This is where we are taking it.
- Next
Push leads to your CRM
Leads flow straight into the tools you already use, and export as a file.
- Next
Lead map
See requests on a map and by date.
- Next
Upload knowledge in the dashboard
Add a file, paste a web address, or forward an e-mail, and it becomes knowledge after a click.
- Next
Open API and webhooks
Connect a website, a booking tool or a CRM with one token.
- Next
Cheaper with every week
Repeat questions answered from approved replies, so the AI is used less.
- Next
Learns from your edits
Your corrections shape the next drafts for your brand.
- Next
Analytics and monthly report
Volume, response times, and how much it handled, in your inbox each month.
- Next
More inboxes
Connect Google and Microsoft inboxes directly.
- Next
Web chat and voice
The same brain on every channel.
- Next
Look things up
Read-only checks of availability or an order, and reading attachments.
// part of the family
Next: working alongside Oper8X
Oper8X is our one-screen view across our products. Linking Emlrun to it is on the roadmap, so that what came in, what waits for your yes and which leads were saved show up there too. Today, all of this lives in Emlrun's own dashboard.
EMLRUN
- What came in
- What waits for your yes
- Which leads were saved
OPER8X · ONE SCREEN
Planned: Emlrun reports here. You ask, approve and steer.
Needs your yes
Software that does the work, not software you feed.
Most e-mail tools give you a better inbox to type into. Emlrun does the work itself: it reads, checks, answers and records, and brings a person in only when a person is needed. It is part of the EzTechTonic family, alongside KLM, our self-learning AI layer.
// questions
Frequently asked questions
Will Emlrun send replies without me seeing them?
Only where you allow it. For each kind of e-mail you choose whether replies are sent automatically, drafted for your approval, or handed to a person. It sends by itself only when every check you set passes, and e-mails it sees as a complaint, a legal or a safety matter are handed to a person. A launch starts draft-first, with every reply waiting for approval until you choose otherwise.
Where do the answers come from?
From your own documents and pages, which we load as its knowledge when we set it up. Drafts are written from them, and a price or figure that is not in them sends the e-mail to a person. If nothing relevant is found, Emlrun does not guess: the e-mail goes to a person.
What if someone tries to trick it?
E-mails that look like they try to change its instructions or pull out private information are flagged and handed to a person, not answered automatically. Emlrun is also tested against a set of such e-mails, and every one of them must go to a person.
Can it work for more than one brand?
Yes. One Emlrun can serve many brands, each with its own knowledge, voice, rules and data. They are kept separate, and the separation is enforced in the database, not only in the screens.
Is there a record of what it did?
Yes. Every step an e-mail goes through is written to an audit log. Replies sent by Emlrun carry the sign-off and AI disclosure line that you set.
How do I get Emlrun?
Emlrun is set up for each business with our team, and there is no self-serve sign-up yet. Tell us about your inbox on the start a project page and we will come back to you.
Want your inbox answered like this?
Tell us about your inbox and what you want it to do.