Inbox Triage
It's 7:30 and eighty-one messages came in overnight. This agent reads each one and makes the call: a new job inquiry routed to the estimator who covers that zip, a supplier confirmation filed against the job, a quote follow-up that's been sitting eleven days pushed up where you'll see it, a vendor pitch sorted to the bottom. Routine replies arrive as drafts. Nothing sends without you.
Yours outright — no required subscription.
Every agent is built for your business — your systems, your approval chain, your way of working. We scope it on the call.
If this sounds too technical, don't worry. We take care of everything for you.
- Private VPS deployment. This agent runs on a server that belongs to you, not a shared cloud tenant. Your mail is read on your own box, and nobody else's messages touch it.
- A technician agent alongside it. Every hire ships as a pair: the Inbox Triage Agent doing the sorting and drafting, and a VPS-technician agent on the same box keeping it patched, backed up, and monitored. You're not hiring one thing, you're hiring a small team of two.
- Communication channel connection. Wired into the shared mailbox where messages actually land, and into wherever your team already talks, so nothing has to be forwarded to it by hand.
- A routing map built from your real mail. We go through weeks of your actual messages and write down the categories, who each one goes to, and what counts as urgent at your company — including the routing rules that currently live in one person's head.
- Security hardening. Private VPN, firewall, and encryption configured on your server before it reads a single message.
- A live walkthrough plus 14 days of priority support. We sit down on a call and show you what it labels, what it drafts, and where it stops. Then two weeks of priority support, to fix the categories your real message mix breaks.
- You own everything — the server, the agent, the code it runs on.
Not sure it fits? Check fit in 90 seconds in the free assessment chat.
Who this is for
- You're a 15–30 person contractor or trade shop where two public addresses dump into one shared mailbox, one person keeps it open all day, and read-versus-unread is the only record of what got handled.
- You run a 12–20 person agency where client email lands in the account lead's personal inbox and copies a general address nobody actually owns. "Did anyone answer this?" is a daily question, and sometimes the answer is no.
- You're a clinic or practice with six to ten providers whose front desk address gets refill requests, prior auths arriving as scanned PDFs, no-show notices, and a rep asking about lunch — and only one of those has a clock running on it.
- You're the owner of a 30-person firm still reading 200 emails a day, because you're on every thread to catch the two that would cost you a client if they got dropped.
How it earns trust
The failure mode of AI in an inbox isn't that it mislabels things — anyone sorting eighty messages before 9am does. It's that you can't tell which ones it got wrong, or why, until a customer tells you. This one bets the other way: every call it makes is stamped, and the pile it calls routine is five minutes of reading.
A daily review of everything it called routine
Low-priority messages sort to the bottom of a pile you can open and read in about five minutes. In week one that's the check: it tells you whether its idea of routine matches yours, before you trust it with anything.
A stamped record on every message
The label it assigned, which rule or past thread it matched, who it routed to, and whether it wrote a draft or pinged a human. The original message stays in the mailbox with its original timestamp — the agent's decision is an added label and a log line, not a move you have to reconstruct.
Misroutes show up as a pattern
When someone keeps dragging the same category from the estimator over to the bookkeeper, the log shows the same wrong call a dozen times, so you fix the rule instead of retraining a person. Because it runs on your server against your mailbox, you pull that log yourself.
The 14-day support window after delivery exists for exactly this, when your real message mix hits it and the first rules need adjusting.
Pairs well with
These share a workflow with this role. Tick any to add them to your setup.
I don't need it answered. I need to know which ones need me.
That's the main job. Every message gets a label and a route, and the ones that need a person go to the top with the reason attached. The routine replies come drafted underneath, and you can ignore that half of it entirely if all you want is the sort. Most people start using the drafts after a few weeks, but the sorting is what changes your morning.
Does it send email as me?
Only for the narrow list of message types you approve during setup, and only after you've watched it handle them. Everything else lands as a draft in a review queue with your name still on the send button. Silence from it means a draft is sitting there waiting on you — it never means a reply already went out.
Will it clear out the junk for me?
No, and that's deliberate. It never deletes, archives, or files a message out of human view — the worst thing a triage system can do is make a message disappear. Low priority means sorted to the bottom of a pile you can still read, not gone. It also never invents a route: if a message doesn't match a rule, it goes to the person who owns the mailbox, not to a guess about who probably handles this.
Stuff falls through the cracks whenever Sandy's out. Does this fix that?
Partly. It replaces read-versus-unread with a written record — what came in, what it was called, who it went to, whether anyone answered — so someone covering the mailbox can see the state of it without having worked there six years. What it can't do is notice what nobody wrote down. If the estimator covering the north side changed this month, it will keep following the old rule until you change the rule.
A quote request sat four days and they went with somebody else. Would it catch that?
An unanswered thread aging past your threshold is exactly the kind of rule it holds better than a person does. Where it's weaker is the quiet emergency: "just checking on that timeline" from a GC about to move a PO reads like routine follow-up, and tone is a bad proxy for stakes. So escalation gets written around who sent it and how long it's been open, not around how urgent it sounds. It also only knows what's in the thread — if a date changed on a phone call, a draft quoting the old one will read as verified when it isn't. That's why drafts get reviewed.
What's covered under the $1,997 setup, and what isn't?
One company, one mailbox as the system of record, one approval chain. Multiple mailboxes with genuinely different rules, or a second company sharing the box, get scoped on the 15-minute call rather than priced as an add-on. Typically about seven days from kickoff to it triaging real mail. It also never changes your escalation path or your response-time commitments: if a message goes to the owner today, it still goes to the owner. It adds a sorting step; it doesn't remove a person from the chain.
Hiring more than one? A department on tap — the subscription puts a build team behind every request, agent after agent.