Work, with the numbers attached
Where a figure was recorded before the work started, it is here with the number it moved to. Where an engagement was never instrumented, the case study says so rather than reaching for an adjective. Most clients are anonymized by descriptor and industry, because much of this work touches internal processes the client would rather not publish.
The method is the same throughout. A process is measured before anything is built, a system is built to remove it, and the result is reported against the number recorded at the start. Where a figure is an estimate rather than a measurement, it says so, and where an engagement was never instrumented at all, the case study says that instead of implying a result it cannot evidence.
The stack varies more than the method does. What decides the shape of an engagement is the process, not the technology, and several of these ran across more than one of the four services.
roughly $77M
political TV ad spend tracked and attributed
Turning FCC filings into roughly $77M of tracked political ad spend
A four-stage LLM pipeline that turns raw FCC contract filings into queryable weekly ad spend, with every dollar traceable to its source document.
a US political advisory practice. Political intelligence and advertising.
- AI agents and LLM systems
- Data and backend engineering
- Workflow automation
Read the case study
90%
lower monthly automation cost, from $500 to $50
From $500 to $50 a month: rebuilding an agency automation stack
Migrating a digital agency off a hosted no-code platform onto self-hosted n8n and Python cut costs 90% and responses from 3 minutes to under 20 seconds.
a digital product agency. Digital agencies and product development.
- Workflow automation
- AI agents and LLM systems
- AI strategy
Read the case study
under 2 minutes
to compile the weekly lead report, down from 5 hours
From 5 hours to under 2 minutes: a legal firm's lead report
Re-engineering a manual weekly lead report with n8n and AI-powered validation cut it from 5 hours to under 2 minutes for more than 12 executives.
a US legal services firm. Legal services.
- Workflow automation
- AI agents and LLM systems
Read the case study
an estimated 75%
fewer manager interruptions, by the client's own estimate
An estimated 75% fewer manager interruptions across 5 locations
A bilingual English and Spanish operations assistant answers 20+ daily staff questions about recipes and SOPs across a 5-location restaurant group.
Padrino's Cuban Cuisine. Hospitality, food and beverage.
- AI agents and LLM systems
- Workflow automation
Read the case study
eight
specialized assistants reachable from one Slack conversation
Eight AI assistants a creative agency talks to in Slack
A master agent reads requests arriving in Slack and delegates them across eight specialized assistants, so nobody routes the work by hand.
Crescendo Creative. Advertising and marketing.
- AI agents and LLM systems
- Workflow automation
Read the case study
2 hours
runtime of a critical data process before the work
From 2 hours to 9 minutes: re-engineering an enterprise data process
Founder experience, not a Datasmarts engagement. Re-engineering the query layer of a critical data process cut its runtime from 2 hours to 9 minutes.
a multi-billion dollar enterprise client. Enterprise software.
- Data and backend engineering
Read the case study
under a minute
agent response time, down from minutes per response
From minutes to under a minute: a guest messaging RAG rebuild
Rebuilding an over-engineered retrieval layer from first principles, and moving ingestion off a monolith, cut AI guest replies from minutes to under a minute.
Cortado, Inc.. Property rental, real estate and travel.
- AI agents and LLM systems
- Data and backend engineering
Read the case study
Your process is probably on this list in some form
The reading that costs a person a day a week, the report nobody wants to compile, the questions that interrupt the same manager. Tell us which one is yours and we will tell you whether it is worth automating.