Insights

AI in business, concretely — costs, compliance, architecture, funding.

• 30 July 2026

  RAG for technical manuals: 7 controls before rollout

  How to deploy RAG on manuals and service procedures: versions, permissions, citations, refusal, golden sets, approval and monitoring.

• 22 July 2026

  API, RPA, RAG or an AI agent: how to choose the right mechanism

  A practical decision tree for choosing API automation, RPA, RAG, an AI agent or no-go based on process, data, risk, error cost and control before a PoC.

• 15 July 2026

  7 Evidence Pack items an auditor or board will ask for

  Which evidence should exist before an AI system reaches production? Seven Evidence Pack items: owner, risk, golden set, evaluations, logs, handoff and rollback.

• 9 July 2026

  AI for handling email queries — a knowledge base, sources and human handoff

  How to automate email query handling with RAG and an AI agent: classification, answers with a source, escalation rules and human handoff. Not an autonomous inbox, but consistent answers under control.

• 9 July 2026

  AI for B2B e-commerce and distribution — where it actually pays off fast

  Where AI makes sense in wholesale and B2B distribution: pre-meeting customer briefs, order-drop alerts, cross-sell, price monitoring and a safe shop bot. Process, data and a first PoC.

• 8 July 2026

  How to choose an AI vendor in 2026: 12 questions before you sign

  12 questions to ask an AI vendor before a PoC, RAG or agent: data, security, the AI Act, evaluations, code ownership and maintenance — with a strong-vs-weak signal matrix.

• 6 July 2026

  AI concierge for hotels — RAG and agents for guests, with clear limits

  How a hotel can use RAG and AI agents for guest service: a multilingual concierge from your own materials, with clear limits and a realistic first PoC.

• 6 July 2026

  AI in insurance — RAG for claims documents, without deciding the payout

  How to use RAG and AI agents to analyze claims documents, policies and procedures — with a source and an audit trail, without automating the payout decision.

• 6 July 2026

  A safe AI PoC in 30 days: a step-by-step implementation plan

  An operational plan for a safe AI PoC in 30 days: pick a process, inventory shadow AI, prepare data, run a PoC on a real slice, evaluations, guardrails, logs and a go/no-go decision.

• 2 July 2026

  22 July and 2 August: who does Article 50 AI Act really apply to?

  The 22 July initial-signatory deadline for the voluntary code has passed; Article 50 duties will apply from 2 August. Check role and use case.

• 2 July 2026

  AI readiness audit: what exactly we check and what you get at the end

  An AI readiness audit step by step: processes, data, shadow AI, AI Act risk classification and an implementation roadmap. What the final report contains.

• 2 July 2026

  The Digital Omnibus and the AI Act: what moved to 2027/2028, and what still applies from 2 August 2026

  Regulation (EU) 2026/1744 is in force from 27 July 2026. Annex III moves to 2 Dec 2027 and Annex I to 2 Aug 2028; most AI Act dates stay.

• 30 June 2026

  Data readiness for RAG — a checklist before you deploy AI

  Before you build RAG, check the data: sources, permissions, document quality, freshness and tests. A data-readiness checklist before you deploy AI.

• 29 June 2026

  AI Act glossary — key terms in plain language

  The AI Act in a nutshell: AI system, prohibited practices, high-risk, GPAI, provider vs deployer, human oversight, KRiBSI. Each term in one sentence.

• 29 June 2026

  AI Act risk classification — a step-by-step method

  How to tell whether your AI system is prohibited, high-risk or minimal under the EU AI Act (Regulation 2024/1689). A step-by-step classification method.

• 29 June 2026

  How much does an AI agent or RAG system cost in Poland in 2026? Cheap chatbot vs production system

  AI price lists start "from a few thousand złoty", but a price without scope says nothing. Cheap chatbot vs a production RAG/agent system — what drives the cost.

• 29 June 2026

  AI: buy off-the-shelf or build your own? A decision rubric

  When to take a ready-made AI tool and when to build your own on your infrastructure. Five questions that settle build vs buy — and why the answer is often a mix.

• 29 June 2026

  AI governance for regulated industries — what it means in practice

  AI governance is not a document but an ongoing function: owner, roles, human oversight, evaluations, data control and audit. What it means in finance, healthcare and manufacturing.

• 29 June 2026

  AI on AWS compliant with GDPR and the AI Act

  How to design AI on AWS with control over Regions, access, logs and retention. Bedrock, Guardrails, PrivateLink, CloudTrail and responsibility.

• 29 June 2026

  Production-grade AI agents — how they differ from a chatbot, and how to keep control

  A chatbot answers; an AI agent acts — it uses tools and reaches into your systems. With autonomy comes risk — how to build and supervise a production agent.

• 29 June 2026

  Evaluating RAG — how do you know it answers accurately and from sources

  How to measure RAG quality: retrieval relevance, answer groundedness, citation correctness and hallucination control. Five metrics and a method.

• 25 June 2026

  mojApteczka as proof: GenAI and RAG in healthcare from idea to production

  How a production GenAI system was built in a high-stakes domain: RAG architecture, hallucination control, inference cost and GDPR plus AI Act compliance. Hard numbers from production.

• 21 June 2026

  AI grants in Poland 2026: FENG, Dig.IT and KFS without the bureaucratic noise

  Got an AI project but the budget is blocking it? Three programmes — FENG / SMART Path, Dig.IT and KFS 2026 — which fits, what qualifies and what does not.

• 21 June 2026

  USD 0.0006 per AI scan: what GenAI really costs in production

  We have our own production number: USD 0.0006 per AI scan. But inference cost is the tip — the real cost of GenAI sits before the model, not in it.

• 21 June 2026

  Guardrails are not an AI policy. They are a technical layer you must test

  Teams confuse AI policy with technical safeguards. Guardrails (e.g. AWS Bedrock Guardrails) help — but without tests, evals and an owner they are decoration.

• 21 June 2026

  RAG on company documents — when a chatbot starts answering from sources

  What RAG is, how it connects AI to your company knowledge base, and how it limits hallucinations — with source citations, on AWS.

• 28 May 2026

  AI Act 2026 — a readiness checklist for the board

  Regulation (EU) 2026/1744 is in force. Most AI Act provisions apply from 2 Aug 2026; full high-risk duties follow in 2027/2028. Board checklist.