How we build AI products, without the hype.
Notes from engineering practice: what it takes to get an AI feature from a demo to production, where such systems break and when they are not needed at all. No round-ups, no invented numbers.
OpenAI Agent Builder is shutting down: how to migrate an agent without breaking it
OpenAI shuts down Agent Builder on 30 November 2026. There are two official paths: the Agents SDK if the agent lives in your application, and ChatGPT Workspace Agents if your staff use it inside ChatGPT. A code export exists, but it does not convert the workflow graph. So migration is mostly verification: capture real inputs before the shutdown and replay them on the new version.
AI agents6 min readAI agents for business: which tasks to delegate
AI agents for business can help with workflows that involve interpreting text, choosing a next step and using several systems. Start with one process and define its result, permitted actions and acceptance checks. The scenarios and selection table below help you choose a pilot.
AI agents5 min readHow much does an AI agent cost?
An AI agent's cost includes development and the ongoing cost of a working process. A model tariff covers only part of that total: integrations, retries, maintenance and human review also matter. Before comparing proposals, define the result you are paying for.
AI agents5 min readAI implementation: plan your first business pilot
A first pilot should answer a specific question: can the system complete the chosen task at acceptable cost and quality? Document the current workflow and acceptance conditions before development. Those inputs support a decision to launch, revise or stop.
AI adoption5 min readRAG vs fine-tuning: what should a knowledge assistant use?
If an assistant must answer from changing documents, start by checking retrieval and sources. If the problem is consistent output format or behaviour, test instructions and then consider fine-tuning. RAG and fine-tuning can work together; the choice depends on the failure you need to fix.
RAG & knowledge base6 min readWhy an AI bot may reply twice: what to check before launch
In our own product, a retry covered the model request, saving the reply and sending a notification. A failure in the last step could restart the whole sequence. A code review caught the risk before release.
AI agents9 min readDify, n8n or custom development for an internal AI agent
Dify is a useful starting point for an AI application with models and knowledge. n8n starts from workflows that connect services and actions. Custom code becomes relevant when access rules, action control or operations exceed what the chosen platform can support. Test the choice on one real workflow.
AI agents7 min readHow to evaluate a RAG system before production
Evaluate RAG with questions people already ask of your documents. For each question, record the correct source, an acceptable answer and when the system should refuse. Measure retrieval and the final answer separately. That tells you whether to fix data, search or the model.
RAG & knowledge base12 min readAI agent platforms in 2026: what a business should choose
Eleven platforms, what each of them charges for, where its ceiling is and what actually works from Russia. Every figure comes from the vendor's own public page, checked on 6 September 2026, with the source named. No rankings and no top-ten lists: the choice depends on what you are building, not on a position in a table.
AI agents10 min readn8n or custom development: where the line runs
n8n covers more than people assume, and we regularly recommend starting there. But a workflow builder has a boundary, and it runs not through scenario complexity but through the cost of a mistake, concurrency and access rights. How to work it out in advance and when writing code becomes the cheaper option.
Process automation9 min readOff-the-shelf chatbot or custom: counting the cost of ownership
An off-the-shelf bot launches in a week and looks ten times cheaper. The difference shows up not in month one but where the bot has to look inside your systems, answer from your documents and hand over to a human honestly. What the two-year bill is made of and how to choose without overpaying either way.
AI chatbots9 min readChunking documents for RAG: why chunks matter more than the model
A bad RAG answer almost always comes from chunking rather than from the model. How to split policies, contracts, scanned PDFs and support tickets, what breaks on a corpus of tens of thousands of long documents, and how to tell that chunking is the culprit.
RAG & knowledge base8 min readHow to update a RAG knowledge base without stopping the system
A demo lives on a static corpus, production on a moving one. Incremental updates instead of full reindexing, swapping the index without a restart, deleting documents, and which metrics reveal retrieval degradation before users report it.
RAG & knowledge base9 min readHow to build an AI agent: what you need beyond the prompt
The prompt and the model are the smaller part of the job. An agent becomes usable when it has tools, explicit authority boundaries, state, a log and cost control. A step-by-step breakdown of what it is made of, how to get it to production and where it breaks.
AI agents8 min readWhat RAG actually is and when a business needs it
RAG means the model answers from your documents rather than from the internet, with a visible source under every claim. What it is made of, where quality is lost in chunking and updates, how to measure that it works and when you do not need it.
RAG & knowledge base