Glossary · AI without the fog
AI without the fog. Explained in plain words
Glossary22 terms
The terms you'll come across with us, explained normally, without technical jargon. You don't need to master this; we do. It's just here so you know what we're talking about.
Glossary
The basics
What AI is and how it actually thinks
- Artificial intelligence (AI)
- Software that handles tasks once reserved for humans: writing, sorting, deciding by patterns. No magic, just a very capable tool.
- Model
- The AI's "brain": a trained program that predicts the most likely output from your input. Different models do different things: text, images, voice.
- LLM (language model)
- A model trained on huge amounts of text that understands and produces language. It powers tools like ChatGPT or Claude.
- Generative AI
- AI that creates something new (text, images, code, voice) instead of just sorting existing data.
- Prompt
- The instruction or question you give the AI. The clearer the prompt, the better the result.
- Token
- A piece of text (roughly a syllable or word) the AI uses to measure input and output. Running costs are usually billed by token count.
- Context window
- How much text the AI can "hold in its head" at once. In a very long conversation it may forget earlier parts.
- Hallucination
- When the AI makes up information that sounds credible but isn't true. That's why we always verify the important stuff.
Glossary
How we work with AI
The tools and methods we build for you
- Automation / workflow
- A chain of steps that run on their own: an email arrives, AI sorts it, creates a task. Saves hours of manual work.
- Agent
- AI that doesn't just answer but takes steps toward a goal, using tools, searching and deciding within clear rules.
- API
- How programs talk to each other. It lets us connect AI to your systems (customer records, website, email) without manual copying.
- Integration
- Connecting AI or automation to the tools you already use, so everything works in one place.
- RAG (answers from your data)
- The AI first finds the answer in your own documents, then explains it. It answers from your data, not the internet.
- Fine-tuning
- Additional training of a model on your data and style so it answers exactly the way you need.
- Vibe coding
- Building apps by describing them in plain language instead of hand-coding, a fast path from idea to a working version.
- MCP
- A standard that lets AI safely reach into your tools and data. Simply put, "USB-C for AI", one connection for many systems.
Glossary
Operations, data & security
How it runs and how we protect your data
- Cloud vs. local
- AI runs either on remote servers (cloud) or on your premises (locally). We choose based on data sensitivity and cost.
- Human oversight
- The AI prepares a draft, a human always has the final word. That's how we keep quality and accountability on sensitive steps.
- GDPR
- European rules for personal data protection. We build solutions to comply with them.
- Anonymization
- Removing or masking personal data so it can be worked with safely.
- Running costs
- AI is usually billed by usage (tokens or calls). We always state an estimate and a cap upfront.
- Ownership
- What we build is yours: code, data and access stay with you. Nothing ties you to us.
Glossary
Found a term that isn't here?
Ask anytime: that's what we're here for. No question is a dumb question.
Start here
15 minutes is enough for you to know whether it makes sense.
You tell us what is bothering you. We tell you whether we can help, roughly what it costs and where I would start in your place. No presentation.
- jakub@transformuj.ai
- Phone
- +420 773 070 090
- Address
- Legerova 1820/39, Praha 2