MERIDIAN CIVIL Variation claims workflow 11 claims/month average
ILLUSTRATIVE COMPOSITE — fictional company; timings follow patterns cited at the foot of the page, flagged as estimates until instrumented
What producing this sheet took: one sit-down with one estimator, reconstructing a claim he'd already
finished — about an hour, at his desk, between deadlines. No site time, no systems access, no observers with clipboards.
This is the entire cost of finding out where a workflow's money actually goes.
85 min — judgment
255 min — fetching, checking, re-typing, formatting
44 min — records
Green — decisions only your estimator can make. Protect it.Rust — work a machine should carry. Automate it.Sand — required records. Streamline it.
The steps, as the estimator told them
Step
Minutes
Pile
The temp test: “could a capable temp with a good checklist do this?”
Dig out site diaries, delivery dockets, survey data from four systems
65
MACHINE
Yes — it's retrieval. The AI can't do it today only because it can't see your systems. Fixable.
Decide claimability: which delays are contractually arguable, which to absorb
30
JUDGMENT
No. Contract knowledge + relationship knowledge. This is the fee.
Draft the narrative (AI-assisted on a personal account, quietly)
40
MACHINE
Mostly — the draft is fine; he rewrites it heavily because no one has defined what “good” looks like.
Price the variation: rates, margins, what the client's QS will accept
35
JUDGMENT
No. Twenty years of knowing where the line is.
Cross-check quantities against drawings; chase two discrepancies
70
MACHINE
Checking is machine work; deciding what to do about discrepancies is the 20-minute judgment inside it.
Final read: “would I sign this in front of the superintendent?”
20
JUDGMENT
No — and this is the check that catches a confident, wrong AI answer. Keep it, name it, write down what it looks for.
Reformat into client template; fix numbering; assemble appendices
80
MACHINE
Yes. Pure assembly. Nobody should be paid senior rates for this.
Register entries, photo log, correspondence trail
44
RECORDS
Required — and mostly auto-fillable from the steps above.
What this means, in money
The judgment core is 85 minutes of a 6.4-hour task — 22%. The other 78% is scaffolding
around it. At 11 claims a month and senior-estimator rates, the scaffolding costs
≈ $14,300/month in this workflow alone — and the judgment, the only part a competitor
can't copy, currently has no written standard, so it can't be taught, checked, or defended.
What we do with this sheet: automate the rust (with your systems connected, so answers stop being generic),
write down the green as Meridian's standard — what a claimable delay looks like, what the sign-off read checks —
with worked examples from your own best claims, and streamline the sand. Timings above are the estimator's
reconstruction; week-3 instrumentation replaces every number with measurement before anything is built.
Not a civil contractor? Swap “variation claim” for your tender response, your inspection report,
your scope letter, your client advice. The proportions move; the three-pile shape doesn't — and the sheet is
produced the same way: one hour, one person, one finished piece of work.
Where the composite's numbers come from — the quiet personal-account AI use is the
national norm (21–30% of employees, Jobs & Skills Australia / Employment Hero); the heavy-rewrite pattern
(“no one has defined good”) is the most common finding in UTS HTI's SME survey; the judgment-to-scaffolding
proportions reflect the coordination-heavy structure of claims work at market senior-estimator rates.
In a live engagement this sheet is built from your workflow, and instrumented before anything is automated.