Every AI pilot report has the same headline: ‘saved the team 12 hours a week.’ It's the easiest number to produce and the least convincing one to put in front of a CFO. Hours saved doesn't tell you whether the work got better, whether the client noticed, or whether that freed-up time actually left the building instead of getting reabsorbed into email. If you're building the business case for Claude inside an Australian firm, hours saved should be the metric you mention last, not first.
Why hours saved is the wrong headline metric
Hours saved assumes the baseline process was error-free and that freed-up time converts directly into cost reduction. Neither assumption usually holds. A paralegal who used to spend four hours drafting a first-pass contract review and now spends one hour with Claude hasn't necessarily saved three hours of payroll. She's more likely spent that time on two additional files, or on the exceptions that used to get rushed at month end. The signal that matters isn't the hour count. It's what happened to the output: did error rates drop, did files move faster through the pipeline, and did the client actually rate the result higher.
Four metrics that actually move a business case forward
Finance teams and boards respond to metrics that map to cost, risk, or revenue. Four in particular hold up under scrutiny:
Error rate: the share of work needing rework, correction, or a second review. A drop from 8% to 2% on invoice coding is a concrete, auditable number a board will trust.
Cycle time: how long a unit of work takes from intake to completion, measured end to end rather than just the AI-touched step.
Net Promoter Score or client satisfaction: whether the person receiving the output rates it higher, not whether the team simply feels faster.
Rework and escalation volume: how many items bounce back to a senior reviewer, a compliance check, or the client for correction.
Each of these is measurable with data most firms already collect: ticketing systems, practice management software, and client surveys. That makes them far easier to defend in front of a partner group than a self-reported time diary.
Taken together, these four numbers describe a fuller picture than hours saved ever could. Error rate and rework volume speak to quality and risk. Cycle time speaks to throughput and capacity. NPS speaks to whether clients can actually feel the difference. A firm that only tracks time saved can end up in the odd position of reporting a successful pilot while nothing about the client experience or the error rate has moved at all.
A worked example from an Australian professional services firm
A 40-person Sydney accounting practice rolled Claude into its BAS review process earlier this year. The hours-saved number looked modest: about six hours a week across the team. The metrics that mattered told a different story. Cycle time on BAS reviews dropped from an average of 3.2 days to 1.6 days during peak lodgement periods. Error rate on first-pass reviews, meaning returns kicked back by a senior accountant for correction, fell from 14% to 5%. That reduction in rework was worth an estimated $68,000 a year in senior staff time that had previously gone to fixing junior work instead of reviewing new client engagements. Client satisfaction on turnaround time, tracked through the firm's existing post-lodgement survey, rose from a score of 32 to 51 in the same quarter.
None of that shows up if you only track hours saved. It shows up when you track what the business already measures for other reasons: reviewer time, rework volume, and a client survey the firm was running anyway.
Building the business case: what to measure before you start
Before rolling Claude out to a team, take a baseline on the four metrics above for at least four to six weeks. Most firms skip this step and then struggle to prove impact three months later because they have no pre-implementation number to compare against. The baseline doesn't need to be sophisticated. A shared spreadsheet tracking rework tickets and cycle time by week, started before go-live, is enough.
It's also worth separating the metrics you'll report externally, to a board, an insurer, or under APRA-adjacent governance requirements if you're in financial services, from the ones you'll use internally to tune the rollout. A regulator or auditor will want error rate and rework volume with an audit trail. A team lead will want cycle time by task type so they know where to focus training.
If you're weighing up a Claude pilot and want a second set of eyes on which metrics actually fit your workflow, a short conversation usually surfaces the two or three that matter most for your context. Book a 30-minute session and we'll work through it against your own numbers rather than a generic template.
