The work
Four engagements across four industries. Different workflows, different interventions, different approaches — each selected because it was right for the problem.
Manufacturing
Invoice Reconciliation
An AP team reconciling 200 supplier invoices per month manually — with a 30–40% overbilling miss rate and no exception tracking. The intervention used Python matching logic, not a language model, because the problem is deterministic.
View case study →
Logistics
Spot Quote Response
A freight brokerage with an 18.4-minute average quote response time. In spot freight, speed wins loads. An LLM drafts the quote email; the broker reviews, adjusts the rate if needed, and sends.
View case study →
E-Commerce
Product Description Generation
A DTC brand producing descriptions for 200+ SKUs per seasonal launch — 30–60 minutes per SKU, 55% Creative Director revision rate. A prior AI tool had failed. The difference was implementation, not technology.
View case study →
FP&A
Variance Commentary
FP&A analysts spending 9 hours per month writing budget-vs-actual commentary from a blank page — every month, for recurring patterns. The close model is structured data; the output format is defined; the problem fits the tool.
View case study →