Professional Services /Document Generation live field guide · 8 min

Evidence Chronology Builder

Turn a pile of 50+ emails, contracts, invoices and file notes into a dated, source-linked chronology your lawyer can actually rely on — in an afternoon, not a fortnight.

Open the live demo

theater/demos/profsvc_chronology.html · sandbox · read-only

Open

FIG. 1

The live demo, running on fabricated data. Open it to step through the full flow — every output is shown for a person to approve before anything happens.

How it would work

Reads every document in the matter, dates each event and links it to its source page, then lays the whole timeline out for a solicitor to check and approve before it goes near a brief or an affidavit.

Input 01

The whole matter file

Every document in the matter — emails with headers intact, contracts and variations, progress claims and invoices, letters, file and meeting notes, texts — as PDFs, Word files, scans or exports.

Agent 02

Dates, attributes, cites

Extracts each dated event, attributes it to a party, orders it on one timeline, links every entry to the document it came from, and flags where an expected document — like a written variation approval — is missing.

Output 03

A draft, working shown

A draft chronology with a clickable source citation on every entry, plus a findings list and gap flags, for a solicitor to check entry-by-entry, correct, and approve before it goes into a brief, affidavit or advice.

Where it works well

It reconstructs the sequence in full, every time, and never asserts an event without a source.

Done by hand it is a day or two of opening files, reading dates off letterheads and typing a table — and the gaps are the first thing missed inside a 53-document bundle.
Best for the solicitor or paralegal building chronologies regularly: construction and Security of Payment disputes, commercial litigation, insurance and employment matters.
Earns its keep above roughly 30–50 documents; the recaptured time goes into deciding what the sequence means, not transcribing it.

The slow, invisible cost in a document-heavy dispute is reconstruction — the facts are real but scattered across an email, an unsigned variation, a progress claim and a file note, and lining them up into one reliable order is hours of low-judgment transcription billed at the wrong rate.

Where it works badly

It is confidently wrong when documents can't be dated from their own four corners — and the draft looks more finished than the dates behind it.

It extracts every dated event, including the trivial ones, and does not know which three of fifty entries actually decide the matter — read as "the answer" rather than "the raw sequence to interrogate", it will mislead you.
If the matter is small enough that one person holds the whole story in their head, the setup costs more than it returns.

The honest test

If a junior solicitor couldn't reliably date and attribute a document by reading its four corners, neither can this — and a timeline of guessed dates is worse than no timeline.

Feed it an undated draft, a forwarded thread three replies deep, or a meeting note with no date in the body, and it will place the event where the text implies and flag the uncertainty — but a chronology built mostly on inferred dates is a liability, not an asset. That is the trap.

What it doesn't do — and shouldn't

It drafts and cites. A solicitor approves each entry. That boundary is load-bearing.

WHAT IT DOES

Surfaces each dated event with a clickable link to its source document

Flags where an expected document — like written variation approval — is "NOT FOUND"

Lays out findings (verbal instruction, proceeding under protest, days overdue) as flags, not conclusions

WHAT IT WON’T

Decide whether a missing approval sinks the claim

Judge which events are material to the matter

Put anything into a brief, affidavit or advice on its own

A chronology can become evidence, and a solicitor carries a paramount duty to the court for the accuracy of what they put forward. An unverified, machine-built timeline has no standing — so nothing leaves review as a finding until a person has checked each entry against its source and approved it. The consequence lands on the practitioner, not the tool.

What your data has to look like

The real, current matter documents — with dates living inside them, not the metadata around them.

Typical readiness

across orgs we see, before the first job

The actual matter documents, not a summary

Needs shaping

Dates that live in the document text

Usual weak point

Legible scans and photos

Usual weak point

One complete, current document set

Needs shaping

The real first job

Most firms have the documents but not the discipline — files live across inboxes, a shared drive and someone's desktop, and the "complete set" is assembled from memory. Getting that capture right — where matter documents live, how they're named, how completeness is assured — is usually the real first job, larger and more valuable than the AI layer on top, and it is a question of how information is filed, not a tool you buy.

Right fit if…

You build chronologies regularly on document-heavy disputes

Matters routinely run above 30–50 sources across many document types

The gaps matter — a missing approval or a claim sat overdue can decide it

Your matter documents can be assembled as a complete, current set

Walk away if…

Most of your matters are small enough to hold in one person's head

The documents are mostly undated drafts and forwarded threads

Files are scattered and the "complete set" is assembled from memory

You want a tool that decides materiality or compliance for you

Open questions

The worried-buyer questions, answered straight

It can mis-date an event, mis-attribute a party, or read an undated draft as if it were executed — which is exactly why every entry carries a clickable link to its source document and nothing is asserted without one. The tool drafts and cites; a solicitor checks each entry against its source and approves it. Treat an unreviewed chronology as a research note, never a finding of fact — and remember a solicitor’s paramount duty is to the court for what they put forward.

Scans and photos run through OCR first, so a faint fax or a phone snap is readable but lower-confidence, and it says so. Undated documents and ambiguous threads are the genuine hard cases: it places them where the text implies and flags the uncertainty rather than inventing a date. The honest test — if a junior solicitor couldn’t reliably date a document from its four corners, neither can this.

No. It removes the hours of opening, dating and transcribing 50-plus documents into a table; it does not form the view about what the chronology means. Deciding which events are material, what a “proceeding under protest” email is worth, and how the gaps bear on the claim stays with the practitioner. The recaptured time goes into that judgment, not off the payroll.

The chronology is only as complete as the documents you give it — it reads what’s in the matter folder when it runs. Discovery is a continuing obligation in Australian litigation, so a late email or a freshly produced variation can shift the picture. Re-run it when the document set changes; don’t treat last week’s chronology as covering documents that arrived since.

It runs against your documents in an environment you control, and the material is not used to train any third-party model. Because these are privileged and often commercially sensitive files, sharing them carelessly with an outside system risks waiving client legal privilege — so the data boundary, who can see the files and how long they’re kept are scoped as part of the engagement, before any matter touches the system. The demo runs entirely on fabricated documents; BuildCorp and SubTech are not real parties.

Document-heavy disputes where the facts are spread across many sources and the order of events matters — construction and Security of Payment disputes, contractual and commercial litigation, insurance and employment matters. It earns its keep above roughly 30–50 documents; below that a paralegal with a notepad is faster.

What it takes to build

3–5 weeks · 4 phases

Reused from template~65%

Bespoke to this skin~35%

stack · Claude · document intake · timeline builder · review UI

What it would cost

Fixed scope, fixed price, fixed dates.

Bite-sized first piece

One contained change, low risk

Pilot build

Most builds land here

Embedded support

Scale on proof

Considering this for your practice?

The honest place to start is a bite-sized first piece — one matter type, one contained change, low risk. Tell us where the reconstruction hurts; we'll play it back, scope it, and show you what's possible.

Book a call How we work