Client case study · 5 months Illustrative

6.5 hours→48 minutes

Time from sample to patient receiving the report, over 5 months.

Elapsed: 5 months
How it was measured: Timestamps from sample registration to delivery confirmation, sampled across 400 reports before and after.
Running alongside: No change to analyser hardware or staffing.
Published with permission: Anonymised at client request

The constraint

Every report typed twice and delivered by hand

Analyser output was read off a screen and typed into a template. The typed report was checked, printed, and handed over at the counter.

Patients called to ask if it was ready, which took more staff time than the typing did.

Results re-keyed from the analyser by hand
Typing errors caught late or not at all
Counter collection only, so patients travelled twice

What we found

The audit, including the unflattering parts

We timed 400 reports end to end.

Median 6.5 hours from sample to patient
3.2% of reports had a transcription discrepancy
Roughly 40 status calls a day

The plan

The 90-day roadmap as it was actually written

Parse the machine, keep the human where the human matters.

Read analyser output directly, no re-keying
Populate the report template automatically
Pathologist verifies and signs off — this stays human
Delivery on WhatsApp once and only once signed

Month by month

What shipped, and what it moved

The line extends as you scroll. Every step below is a real month, including the one where nothing happened.

01Month 1

Baseline

Median 6.5 hours, 3.2% transcription discrepancies, 40 status calls a day.

Baseline: 390 min

02Month 2

Analyser parsed

Results populate the template directly. Re-keying stops.

-23%

03Month 3

Template and formatting

Report layout generated rather than assembled.

-40%

04Month 4

Sign-off gate

After the incident below, delivery cannot fire without explicit verification.

-47%

05Month 5

WhatsApp delivery

Patient receives the signed report without travelling back.

-49%

What went wrong

The part most case studies leave out

We automated delivery before the verification step was locked down, and two unverified drafts reached patients in week three. Delivery is now gated on an explicit pathologist sign-off and cannot fire without it.

This is the one we got wrong, and it mattered.

We enabled delivery before the verification gate was locked. In week three, two unverified drafts reached patients. Nothing clinically incorrect went out, but it could have.

Delivery is now hard-gated on an explicit pathologist sign-off with an audit trail. In a clinical workflow the automation must never be able to outrun the human check, and we should have built that constraint first rather than second.

The numbers

Where it landed, and how we know

Time from sample to patient receiving the report: 6.5 hours → 48 minutes over 5 months. Timestamps from sample registration to delivery confirmation, sampled across 400 reports before and after.

Timestamps from registration to delivery confirmation, sampled across 400 reports before and after.

Sample to patient: 6.5 hours to 48 minutes
Transcription discrepancies: 3.2% to zero
Status calls: roughly 40 a day to under 5

Results reflect this client's market, licence status and starting position. We publish the method alongside the number so you can judge whether it transfers to yours.

In hindsight

What we'd do differently next time

Build the sign-off gate before the delivery path, not after
Run a fortnight in shadow mode before anything reaches a patient
Give the pathologist a one-tap reject that halts delivery instantly

Before you ask us to do this for you

Does this actually apply to your situation?

This worked because of specific conditions. If yours don't match, say so on the first call and we'll tell you what would change.

Market: Diagnostics, pathology and imaging where output is machine-generated and human-verified.
Starting point: An analyser or system that can export results in any structured form.
Timeline: Four to six months. Clinical workflows should not be rushed.

Start with the audit. It's free and it's specific.

Send us a URL. You'll get back the technical issues, your AI-search readiness score, what competitors are doing that you aren't, and what we'd fix in the first ninety days.

Get a free audit → WhatsApp us instead