Work in practice
How I implemented corporate AI for 15 employees and a continuous customer flow
A client implementation with employee memory, company knowledge, live product data, Slack and WhatsApp on a $921 monthly infrastructure and API budget.
Work, people, and processes
Denis Ostapenko
I bring order to day-to-day work.
My background is in commercial cleaning, industrial production, and logistics. I work out how a process runs, agree on responsibilities with the people involved, and make the work easier to follow. AI is one of the tools I use for that.
Operations · People · Practical automation
Selected work
All work and projects ↗A client implementation, a book, and a learning project.
Books
The Word Machine
How language models work, and what it takes to put them to work. AI Academy is a separate set of companion lessons.
Courses and learning apps
AI Academy
A course on how language models and the systems around them work. From the basic mechanism to RAG, infrastructure, and implementation.
Explore by subject
Understand the work. Agree on the next step.
When work is scattered, I start with the process and the people doing it: what needs to happen, who is responsible, and where a task gets stuck. Clear agreements make it possible to follow through calmly.
- Make the process clear
- Understand the sequence, the decisions, and the handoffs between people.
- Talk it through
- Listen, explain what is needed, and agree on responsibilities and a workable next step.
- Automate what makes sense
- Once the process is understood, use software and AI where they help with the actual work.
Books I've written
The bookshelf ↗The Word Machine
How language models work, and what it takes to put them to work. AI Academy is a separate set of companion lessons.
About the book ↗
Your App Has to Work
The work between an AI prototype and an application people can use. A free online guide accompanies the book.
About the book ↗
AI Personas
What has to persist for an AI persona’s character, memory and choices to hold together over time? I follow that question through current research and the systems we can build around a model.
About the book ↗From the notebook
All professional notes ↗How I implemented corporate AI for 15 employees and a continuous customer flow
A client implementation with employee memory, company knowledge, live product data, Slack and WhatsApp on a $921 monthly infrastructure and API budget.
The Agent Said It Was Done. I Still Need to See the Result.
How I check an agent's result, contain mistakes and decide where human control belongs.
Code Stopped Being the Bottleneck. Management Took Its Place.
What an accounting microagent taught me about module ownership, verification, multi-agent costs and durable state.
Elsewhere
My professional profiles and a way to get in touch.