Shovel digs out how work actually gets done before deciding what to automate with AI
jdjohnson · x · 2026-09-08
Mostly Serious launched Shovel, a consultancy addressing why companies that buy AI tools end up with three power users and lots of rewritten emails: the workflows that would actually pay off were never written down.
Core thesis: "A license is not a workflow." How a company actually works lives in people's heads — steps, exceptions, and judgment calls must be dug out before automation. The method runs four phases, one workflow at a time:
- 01 Decide if it's worth automating: can the workflow be defined end-to-end, is the data accessible, can AI use it, is ROI justified? Deliverable is a decision brief at $3,500, about two weeks.
- Later phases dig out how work gets done; every engagement yields one redesigned workflow plus a working memory of the business that speeds up the next one.
Target: manufacturing, engineering, healthcare, and finance companies whose most valuable workflows have never been documented. The deliverable is the workflow; the moat is the memory.
More from coding & agent
- ReCache cuts agentic KV memory 92.43% and speeds first token 3.655x with minimal accuracy loss — techNmak · 2026-09-08
- ReCache reuses tool-schema KV states, cutting agent memory 92% with little accuracy loss — techNmak · 2026-09-08
- Weaviate's year of agent memory: why store-everything RAG breaks three ways — victorialslocum · 2026-09-08
- A simple Next.js app built 2 months ago already has 48 vulnerabilities — haydendevs · 2026-09-08
- GPT-6 Astra 'feels like AGI': $1,000 hands-on across real client work, gaming and computer control — PrajwalTomar_ · 2026-09-08
- Google open-sources Artemis, an Android-automation agent scoring 99%+ on AndroidWorld — Shruti_0810 · 2026-09-08