Practical guides. Honest about limits.
Practical guides, checklists, and worked examples for people deciding how to use AI in a business workflow. Each guide is written by the co-founders, dated, sourced, and honest about its limits.
What is an AI agent?
A plain-English definition of AI agents for business owners: what they do, how they differ from chatbots and automation, where they fail, and how to start.
Read the guide Checklist / 12 September 2026AI agent pilot checklist
Twelve decisions for a first AI agent pilot: task, owner, inputs, permissions, stop rules, test cases, acceptance criteria, cost per task, and handoff.
Read the guide Design / 12 September 2026Human-in-the-loop AI workflows
Where people should stay in an AI workflow: five review patterns, a decision table by risk and reversibility, examples, and how to measure review burden.
Read the guide Checklist / 12 September 2026AI agent security checklist
Twenty questions to settle before an AI agent touches business systems: permissions, data flows, prompt injection, actions, logging, and incident ownership.
Read the guide Method / 12 September 2026AI workflow cost per accepted task
How to calculate what an AI workflow costs to run: model calls, tools, infrastructure, retries, and review labor divided by accepted results.
Read the guide Worked example / 11 September 2026Duplicate invoice checks
See how duplicate invoice checks behave on a synthetic test set. Inspect inputs, comparison rules, review decisions, limitations, and runnable examples.
Read the guideCopy what
we use.
The brief, the acceptance worksheet, and the cost calculator are the working documents behind the guides. They are free to copy into your own files.
Before we begin.
Who writes these guides?
Andrew Matusevich and Seva Rybakov, the co-founders. Each guide records its publication date, its sources, and its limits. Worked examples use synthetic data and say so.
Can we reuse the checklists and templates?
Yes. The brief, the acceptance worksheet, and the checklists are meant to be copied into your own documents. A link back is appreciated but not required.
Are the examples real client results?
No. Examples are illustrative or run on synthetic inputs. They show a method and its limits, not a measured outcome on client data.