Code Sorcery
Case studies

What changed, measured against what it was before.

Every number here is a delta against a baseline we measured ourselves in the first two weeks.

Logistics · 400 SKUs

A supply chain that stopped guessing

Before

A planner rebuilt a reorder spreadsheet every Monday from four exports. It took most of a day, it was stale by Wednesday, and stockouts were absorbed as a cost of doing business.

After

A demand model runs nightly against live sales and lead times, drafts the purchase orders, and flags only the lines where its own confidence is low. The planner now reviews exceptions instead of rebuilding the sheet.

28%
lower cost to serve
35%
faster to deliver
99.4%
inventory accuracy
Demand forecastingERP integrationException routing
B2B SaaS · 12k tickets / mo

Seventy percent of tickets, answered on arrival

Before

Tier-one support answered the same forty questions all day. Median first response was just over four hours, and it got worse every time they launched anything.

After

An agent grounded in their own help centre and changelog answers the repetitive majority with a citation on every reply. Anything it cannot source goes to a person with the history already summarised.

70%
handled without a person
9 min
median first response
4.6/5
satisfaction, up from 4.1
Retrieval agentCitation enforcementHuman handover
Insurance · claims intake

Claims intake that reads its own post

Before

Every claim arrived as an email with attachments in no particular format. Two people spent their mornings opening them, typing the fields into the claims system, and chasing what was missing.

After

Attachments are parsed on arrival, fields are extracted with a confidence score, and anything ambiguous is queued for a human with the source document highlighted. Nobody re-types anything.

11 hrs
returned per person per week
94%
fields extracted unattended
2 days
faster to first decision
Document extractionConfidence thresholdsReview queue
Your turn

What would yours say?

Bring the process that eats your week. We will tell you what the before-and-after would plausibly look like, and whether it is worth doing at all.