This is a machine-readable static content snapshot of the Aletheia on Symplast executive briefing page, provided so that web crawlers and scrapers can index the site's full text without executing JavaScript. The interactive version is at the site root.
Volume Report Triage — Executive Briefing
Aletheia on Symplast is an intelligence fusion capability. Deployed on your infrastructure, behind your accreditation boundary. Australian-owned, sovereign-built. Symplast is patent-pending technology of Psithur Holdings Pty Ltd.
Volume Report Triage · Aletheia on Symplast
It's like standing under a waterfall, trying to catch it all in a teacup.
— Director, Intelligence Triage
01 — The Problem
The problem isn't only the volume. You can't catch the whole waterfall. But right now nobody's choosing which drops to catch either. You are catching whatever fell first.
Hundreds of reports land every day. They get read in the order they arrived. Nobody chose that order. The queue chose it.
01 — The Problem · The Queue
Reading in arrival order isn't triage. It's just what happens.
Here is a day's inbound. An ID, a time, and a subject line someone typed at registration. Nothing on the screen says which of these matters. So the analyst starts at the top.
Eight reports, in the order they were filed. The two-paragraph tip at 13:47 looks like the weakest thing on the list. It is also the seventh thing anyone will read.
When scored by consequence, RPT-04223 moves to the front. It names nobody on a watchlist, so no rule would have caught it. It scores highest because of what it touches: an address you hold under an active priority, a vehicle two steps from a live target, and the third mention of the same associate in 90 days. In arrival order, it was seventh.
02 — The Cost
None of them show up as an incident. They show up later — in a review, in a coronial finding, in a question about what you held and when you held it.
The urgent report and the routine one wait the same length of time, because nothing has told anyone which is which. Speed goes to whatever arrived first.
A name here, a vehicle there, a payment three months later. Each looks minor on its own, so each is treated as minor. The risk you act on today was built out of reporting you already had.
Asked why one report was actioned in a day and another sat for three weeks, the honest answer is that it arrived first. That is hard to say to a court, and impossible to fix, because nothing was written down.
The longer a high-consequence report sits unread, the more it weighs on you. Decisions made on arrival order are decisions nobody actually made — and the risk of that lands on the analyst, not the queue.
03 — What It Takes
Each is common on its own, and most centres already have something that does one of them. The queue only moves when all four work together, on the report's first pass.
Reports are written by people, in whatever format and language their agency uses. Names, vehicles, accounts, places and events have to come out of that writing as it lands — including the ones nobody planned for.
A report is rarely dangerous on its own. It is dangerous because of what it connects to — a watchlist, a live operation, a fragment filed months ago. Those connections sit in a graph, or they don't exist.
What counts as important is your judgement, and it changes with the mission. You have to be able to change the weighting yourself, this week, without raising a ticket with a vendor.
If an analyst can't see why a report scored high, they will open it and check — and you are back to reading in order. Every factor has to point at the sentence, page and document behind it, and still do that in two years.
Three out of four gets you a ranked list nobody trusts and everybody double-checks. All four have to work at once, on one system, inside your boundary.
04 — Why It Persists
Not because these tools are bad. Each does its own job. Each one stops at a different test.
Heading: A rule only fires on what someone already wrote down.
What it does: Flags reports containing words on a list — names, addresses, offence codes. Cheap, fast, and every centre has one.
Where it stops: The report that matters most usually mentions nobody on the list. It's a new name, two steps from a live target, described in words no rule was written for. A list kept by hand also ages faster than the reporting does.
Stops at: READ · CONNECT. (Passes: WEIGH, SHOW.)
Heading: Search answers questions. Triage is not knowing what to ask.
What it does: Indexes everything you hold so an analyst who knows what they want can find it.
Where it stops: The report nobody has read is the one nobody thinks to search for. Search also ranks by how closely a document matches your words, which is a different thing from how much it should change what you do next.
Stops at: CONNECT · WEIGH. (Passes: READ, SHOW.)
Heading: It counts the queue. It doesn't read it.
What it does: Records what arrived, what's open, what's overdue and who holds it. You need this.
Where it stops: Priority is whatever someone picked from a dropdown at registration — usually before anyone read the material, and rarely changed when later reporting made it matter more. It shows you the queue is growing. It won't tell you which end to work from.
Stops at: READ · CONNECT · WEIGH. (Passes: SHOW.)
Heading: It judges the report against itself.
What it does: A model reads each report and returns a summary, sometimes a risk rating. Quick to stand up, and genuinely useful for reading load.
Where it stops: It only sees the one report. The thin tip that connects to a live network reads thin, so it scores thin. And a rating with no sources behind it gets checked by hand — which is the work you were trying to skip.
Stops at: CONNECT · WEIGH · SHOW. (Passes: READ.)
Heading: The model has to be agreed before the first report is scored.
What it does: Maps the domain first — entities, types, relationships — then loads reporting into that structure. Done well, it holds the picture and the audit trail.
Where it stops: Agreeing the model is a programme: workshops, sign-off, migration, all based on predicting what future reporting will say. Anything that doesn't fit gets forced into the nearest box or dropped — and the words that matter are usually the ones nobody had a box for. Until it's finished, the queue is still worked in arrival order.
Stops at: READ — anything the model didn't predict. (Passes: CONNECT, WEIGH, SHOW.)
05 — How It Works
Symplast builds and holds the graph. Aletheia is the layer that weighs each new report against it. Five stages. None of them hidden. No score without its sources.
The report arrives in whatever form you receive it. OCR handles mixed formats. Page numbers, headings and the exact location of every passage are kept.
People, companies, vehicles, accounts, places and events are pulled out of the text. The categories come from what the reporting actually says, not from a schema agreed in advance. Each one is tied to the sentence it came from.
Is this the same person you already hold? Several independent signals vote on the answer. Confidence is measured, not assumed. The analyst can merge, split or overrule.
The report is weighed against the graph: your active priorities, watchlist hits, how close it sits to known networks, and timing. 17 graph algorithms do the work. You set the weightings.
The queue reorders. Every position comes with its reasons: the score, the factors that made it, and the document, page and line behind each one.
Categories come out of the reporting instead of a committee's guess about it. A report using words nobody had a field for still gets read, sorted and scored — on day one, not after an eighteen-month schema project.
Graph features bolted onto a document store can't answer multi-step questions, and that's exactly where risk hides: the tip that means nothing on its own and a great deal two steps out. Symplast is a graph at its core, so a new report is scored against everything, not against itself.
The model proposes, the system checks, the analyst decides — confirming, overruling, changing the weights. Every fact links back to the passage it came from, and every change goes into a ledger nobody can edit. Ask what the system held to be true last March and you get the answer, not an estimate.
Aletheia on Symplast — against the four tests: READ — YES. CONNECT — YES. WEIGH — YES. SHOW — YES.
06 — One Score, Explained
The two-paragraph tip, source not named, filed seventh. Nothing in it stands out. What it touches does.
Two-paragraph tip, source not named. Score: 95. Band: HIGH.
No name in it is on any list, so no rule would have fired. On its own it's a thin document filed mid-afternoon. Against the graph, it's the first report to put two separate holdings in the same place in the same week.
Every line opens the sentence behind it. An analyst can accept the ranking, argue with it, or change the weights — and still explain it two years from now.
07 — Ask This
They are the questions that separate a system that moves the queue from one that only describes it.
If the answer starts with a modelling phase, your queue stays in arrival order until that phase ends — and after it, for anything the model didn't predict.
Ask to see a thin document score high because of what it touches. Then ask how many steps out the system looks before it stops.
Priorities move. If re-weighting is a change request, your ranking reflects last quarter — and people will quietly go back to working from the top of the list.
Anything less is a number your analyst has to verify by reading the report. It's also hard to put in front of a court.
Triage runs on your most sensitive material, before anyone has assessed it. Where that happens is not a procurement detail.
If any answer is no, your queue is still in arrival order. It's just better documented.
08 — Deployment
Aletheia runs on Symplast, inside your centre. Your infrastructure, your control, behind your accreditation boundary.
Air-gapped, on-premises, or sovereign cloud — your choice. No foreign tenancy. Your data. Your decisions.
Connects to your case management, records and partner systems through standard data exchange. Nothing is replaced. The queue those systems show you is scored before anyone works it, with access and disclosure controls built in.
There's no modelling phase to fund first. Point it at a month you have already worked, and compare the reordered queue against the order you actually read it in.
Deployment modes: Air-gapped, On-premises, Sovereign cloud. All inside your accreditation boundary.
Reports get read in arrival order because, until something has read them and connected them, there is no other order to use. Aletheia reads them as they land and puts the queue in the order the mission needs. It is sovereign Australian capability — the code, the data and the decisions stay onshore, under Australian law. Built here, by Australian-owned companies whose founders came out of government, law enforcement and national security.
Australian-owned. Australian-built. Informed by lived experience.
This page is about one problem. Aletheia does more — see the full capability at https://aletheia.pticapability.com
Symplast is patent-pending technology of Psithur Holdings Pty Ltd.