Engagement Intake & Routing Desk
Reads each inbound enquiry, extracts the client and matter details, classifies the request, runs a conflict check, drafts a holding reply, and routes it to the right team — for a person to review before anything is sent.
It triages every new enquiry the moment it lands, surfaces the matter type, urgency, conflict status and route, and drafts a holding reply — then holds everything for your sign-off.
Where this works well
The slow, invisible cost this makes visible is the gap between when an enquiry lands and when the right person actually sees it. New work arrives by email, web form, phone note and attachment, then gets re-keyed across the CRM, the practice system and someone's task list — and while that happens, a time-sensitive matter (a dispute with mediation booked for next Thursday) sits in a shared inbox behind a postal-address change. This reads every enquiry the moment it arrives, pulls out the client and matter detail, classifies intent and urgency, and proposes a route, so the urgent matter surfaces to the top instead of waiting its turn.
It earns its keep most for a multi-partner advisory, accounting or legal firm with a shared intake inbox and several service lines — forensic and disputes, tax advisory, SMSF, audit, business valuation — where the cost of mis-routing or a slow first response is real. The role that benefits most is the intake coordinator or client services lead who currently triages by hand: the repetitive reading and re-keying moves off their desk, and their attention moves to the cases that need a person.
Where it works badly
It is confidently weakest where your reference data is wrong rather than missing. If the conflict register is out of date, the check will come back clean on a matter that is in fact conflicted — the one failure mode that actually carries consequence. The tool can only match against what the register holds; it cannot know about the engagement that was agreed verbally last week and never recorded. Before you trust the conflict beat, run the honest test: pick five recent matters and ask whether the opposing parties and related entities were captured in the register on the day the enquiry came in. If the answer is "mostly, eventually," the register is the first job, not the AI.
It is also poor value for a small firm where one person already sees every enquiry within the hour. If your intake is low-volume and high-context — a handful of referrals a week, each understood the moment it lands — the triage layer adds ceremony without recapturing meaningful time. And for genuinely novel enquiries that don't fit your service-line taxonomy, it will classify with low confidence and route to a human, which is correct but not labour-saving.
What it doesn't do — and shouldn't
It surfaces; a person decides. It proposes a matter type, an urgency, a route and a drafted holding reply, and it flags a conflict — but it does not send the reply, open the engagement, or clear the conflict. Those are decisions with professional-conduct weight. Whether to act once a conflict is flagged is a judgement reserved to a partner under APES 110 §310 for accountants, or Rules 10–12 of the Australian Solicitors' Conduct Rules for legal practices; it is not a setting in a workflow tool. Costs disclosure — required in writing under the Legal Profession Uniform Law once costs are likely to exceed $750 — is similarly a human obligation the tool can prompt but never discharge.
That boundary is deliberate. The point of holding the reply and escalating the conflict, rather than auto-responding, is that the consequence of getting it wrong (acting against an existing client, mis-stating what the firm will charge) lands on the firm and the individual practitioner, not the software. The tool makes the first response fast and the decision visible; the decision stays with the person accountable for it.
What your data has to look like for this to work
Four things need to be in good shape, and most firms have only some of them. The conflict register and active-client list must be current and structured — opposing parties and related entities captured against live matters, not buried in matter narratives — because that is what the conflict check reads. The service-line taxonomy needs to be a real, agreed list (forensic and disputes, tax advisory, SMSF, audit, valuation) that maps cleanly to who owns each, not a folder structure nobody trusts. The routing map needs an actual owner per service line. And your engagement terms — first-response SLA, conflict-check policy — need to exist as referenceable text, not tribal knowledge.
In practice, the conflict register is the one that's usually weakest, and fixing how conflicts get captured at the point a matter is agreed is typically the real first job — bigger and more valuable than the AI layer on top of it. It is rarely a matter of buying a tool; it is a matter of how information is recorded when work comes in. That capture work is what we help with, and it pays off whether or not the triage layer ever goes live.
Could it wave through an enquiry it should have stopped — like one with a conflict?
It is built to stop, not to wave through. Before drafting anything beyond a holding reply, it checks the new party against your conflict register and flags a match — in the demo, the opposing party is an active audit client (Matter #2024-118), so the reply is held and the matter escalated to the engagement partner. It never clears a conflict itself; a person makes that call under APES 110 §310 or the Australian Solicitors' Conduct Rules. If the register is incomplete, the check is only as good as the register, which is why getting that data right is part of the work.
What happens when an enquiry is vague, half-written, or an attachment we can't read?
It classifies on what it can actually extract and shows its confidence, rather than guessing a clean answer. A low-confidence or unreadable enquiry is routed to a person to triage instead of being auto-routed to a team. You will get more 'needs a human' outcomes early on; that is the honest behaviour, not a fault.
Does this replace our intake coordinator or client services team?
No. It does the repetitive first pass — reading, extracting, classifying, conflict-checking and drafting — so your coordinator spends their time on judgement: confirming the route, handling the awkward cases, and deciding what we take on. The capacity it frees goes back into client-facing work, not out of the team.
How current does the data behind it need to be?
The conflict register and the active-client list are the parts that must be current to the day — a stale register is the one failure that matters, because it can miss a live conflict. The routing map and service-line taxonomy can be reviewed monthly. If your client and matter records lag reality, fix that before switching anything on.
Where does the enquiry data go, and who can see it?
It runs inside your own environment — your Microsoft 365 tenancy, your CRM or practice system — not a public chatbot. Enquiry content, client identities and conflict-register data stay within your firm's systems and existing access controls. We scope what the model can read and what it writes back during setup, so privileged or commercially sensitive matter detail is not exposed beyond where it already sits.
Estimated build: 3–4 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.