Skip to content

Work

The problem, the build, and what changed

Every engagement below follows the same shape: find where the hours go, build the smallest thing that removes them, then measure whether it worked.

Illustrative examples. The engagements marked “Example” are representative of the work, not named clients. They will be replaced with real, approved case studies as engagements complete.

Regional wholesale distributor

Distribution · 80 staff

Example

The problem

Three people spent most of each week keying supplier invoices into the ERP. Month-end close slipped because of the backlog, and coding errors were caught weeks later.

What we built

An extraction pipeline that reads incoming invoice PDFs, matches them to purchase orders, and posts them to the ERP. Anything below a confidence threshold or outside tolerance goes to a review queue instead of posting.

The part that mattered was the exception queue. We stopped finding coding errors in the audit.

Finance Director

What changed

94%

invoices posted without a human touch

6 days

cut from month-end close

2.5 FTE

moved off data entry to supplier work

B2B software company

SaaS · 40 staff

Example

The problem

Support volume doubled in a year. First-response times drifted past a day, and the same twenty questions consumed most of the queue.

What we built

An agent that drafts a reply from the knowledge base and past resolved tickets, attaches the sources it used, and leaves it for an agent to send. Nothing goes out unreviewed.

What changed

70%

of tickets answered from a draft with light edits

4h → 40m

median first response

0

extra support hires that year

Professional services firm

Consulting · 25 staff

Example

The problem

Utilisation and pipeline numbers lived in four systems and one much-argued-about spreadsheet. Partners made staffing calls on figures that were a week old.

What we built

Pipelines from the CRM, timesheet and finance systems into one warehouse, with agreed metric definitions and a dashboard that refreshes nightly. The weekly partner pack builds itself.

What changed

1 source

for numbers previously spread across five

8 hrs/wk

of manual reporting removed

Nightly

refresh instead of weekly

Your operation probably has a version of one of these

Tell us the symptom and we’ll tell you, honestly, whether there’s an automation worth building behind it.