Inbox Triage
Turns a chaotic shared inbox into a sorted, urgency-ranked queue with drafted replies — so the morning starts with the work, not the sorting.
Reads each inbound message overnight, tags it by intent and urgency, ranks the queue, drafts a cited reply, and flags anything sensitive for a person to handle.
Where this works well
The slow, invisible problem this makes visible is sorting cost. In most professional-services firms — a legal practice, an accounting office, a consultancy — the shared inbox is the front door, and someone spends the first part of every day reading everything once just to decide what matters. That triage is real work, it is repeated daily, and it produces nothing except an ordered list. This app does that first pass: it reads each inbound message, tags it by intent (a billing query, a booking change, an access problem, a complaint), assigns an urgency, suggests a route, and re-ranks the queue so the genuinely urgent message is at the top instead of fifth from the bottom.
It earns its keep where inbound volume is high and repetitive, and where the cost of a slow reply is real — a client who waits a day for a simple answer, or an urgent matter that sat unread under twenty routine ones. The person who benefits most is whoever currently owns the shared inbox: client services, a practice manager, a small-firm principal doing their own triage at 8am. They start the day on a sorted queue with drafts prepared, not on a wall of unread.
Where it works badly
It is confidently weakest on messages that don't say what they mean. A complaint phrased politely, a serious issue buried in paragraph four of a forwarded chain, sarcasm, or a client who downplays something urgent — these are exactly the cases where a wrong tag does the most damage, and where the tool's confidence score is doing you the favour of admitting it isn't sure. If you treat the urgency tag as gospel and stop reading, the one message that was mis-ranked is the one that hurts.
It is also poor value where your inbound traffic is low-volume and high-stakes — a handful of bespoke client matters a day, each needing a partner's full attention. There is little sorting to recapture there, and the judgement the tool can't make is most of the work. The honest test: if you can describe most of your inbound mail as a short list of recurring types (booking, billing, access, general enquiry), this pays off. If almost every message is unique and consequential, it doesn't yet.
What it doesn't do — and shouldn't
It sorts and drafts; it does not send, and it does not decide what a sensitive message deserves. The line is deliberate. When a message reads as a complaint, a distressed client, or a request for actual professional advice, the tool produces no send-ready draft — it routes the item to a named person to handle directly and says why. That boundary matters here because a complaint mishandled at the inbox can become a regulatory or professional-conduct matter, and advice given by autocomplete is advice your firm is answerable for.
So the human stays on three decisions the tool only assists: whether a drafted reply is right and appropriately worded, whether an unusual message has been understood at all, and what to do with anything sensitive. The tool surfaces, ranks, and drafts. A person approves, edits, reassigns, or escalates. We accelerate the sorting; we do not replace the judgement.
What your data has to look like for this to work
Three things have to be in reasonable shape, and most firms have one of them. First, the inbound mail has to actually arrive in one place the tool can read — a shared inbox or routed web form, not five personal mailboxes nobody else can see. Second, the routes have to mean something: "Support", "Finance", "Reception", "Principal" only help if those are real destinations a message can be handed to. Third, and most often missing, the reply drafts need a source to cite — a current help-centre, a fee schedule, a library of standard responses. A draft is only as accurate as the reference behind it, and a confidently out-of-date answer is worse than no answer.
This is usually the real first job, and it is rarely about buying a tool. It is about how information is captured and kept current: consolidating the inbox, writing down the response library that lives in people's heads, agreeing what "urgent" and "sensitive" mean for your firm. That groundwork is bigger and more durable than the AI layer on top of it, and it is the part we help with. Client correspondence is also personal information under the Australian Privacy Principles, and for many firms is bound by confidentiality or legal-professional-privilege duties — so where the data goes and what is retained is a design decision we make with you, not after.
Could it send a wrong or tone-deaf reply to a client without me seeing it?
No. The tool never sends. It drafts a suggested reply and shows you why it classified the message the way it did, but every draft sits behind an Approve, Edit, Reassign, or Escalate step that a person clicks. Anything it reads as sensitive — a complaint, a distressed client, a request for advice — is held with no send-ready draft at all and routed to a named person.
Our inbox is a mess — mixed threads, forwarded chains, no subject lines. Will this still work?
It will tag the easy, repetitive traffic well (booking changes, hours, address updates, standard billing queries) and that is where most of the volume and most of the wasted time sits. Long forwarded chains, sarcasm, and messages that bury the real request in paragraph four are exactly where its confidence drops — and it shows you that confidence so you can see when to trust the tag and when to read it yourself.
Does this replace our client services or reception person?
No. It replaces the sorting, not the relationship. The person who used to spend the first hour triaging now starts on the ranked queue with drafts already prepared, and spends their judgement on the replies that actually need a human — the complaint, the upset client, the unusual request. Capacity moves to higher-value work; the role stays.
How current does the data behind the replies need to be?
The draft replies cite a source — your help-centre article, fee schedule, or standard-response library — so they are only as right as that source. If your published hours, fees, or policies are out of date, the drafts will confidently repeat the stale version. Keeping that reference content current is part of making this work, and usually the first thing we check.
Where does our client correspondence go, and who can see it?
Client emails are personal information under the Australian Privacy Principles, and for many firms also covered by confidentiality or privilege obligations. The tool reads from your own inbox within your control and the classification happens against your rules; we scope what is sent to any model, what is retained, and what stays in your tenancy as part of the build. This is a privacy and confidentiality design decision, made with you, not an afterthought.
Estimated build: 2 weeks. Most of it is template work we've already done.
Fixed scope, fixed price, fixed dates.
The cost band reflects the engagement shape, not a per-feature line item. We work on fixed scope, fixed price, fixed dates — see the services catalogue for what falls inside each band.
Considering this for your org?
The honest place to start is a bite-sized first piece — one contained change, low risk. Tell us where it hurts; we’ll play it back, scope it, and show you what’s possible.