Glossary
Terms that actually come up in projects.
Short, clear and free of jargon. Use the search or jump straight to a letter.
33 terms in total
33 terms
A3 terms
- AI agent
- A program that completes a task independently across several steps – reading a request, checking data and preparing an answer. Unlike a simple chatbot, an agent can use tools and systems.
- API
- A defined interface two programs use to exchange data. Without APIs, every piece of information would have to be moved between systems by hand.
- Automation
- A recurring work step is handled by software instead of a person. The goal is less effort with fewer mistakes at the same time.
C2 terms
- Chatbot
- A conversational interface that answers questions or accepts tasks – today usually based on a language model and your own knowledge base.
- CRM
- Customer relationship management: the system holding contacts, quotes and opportunities. Usually the anchor for sales automation.
D3 terms
- Dashboard
- An overview screen that brings the key metrics from different sources together – so decisions are not based on gut feeling.
- Data migration
- The controlled move of data from a legacy system into a new one – including cleanup, field mapping and completeness checks.
- Digital assistant
- Software that takes over routine communication and back-office work: pre-qualifying enquiries, coordinating appointments, capturing data, providing information.
E2 terms
- Embedding
- Converting text into numeric vectors so that similar content ends up mathematically close together. The basis for semantic search and RAG.
- ERP
- Enterprise resource planning: the leading system for orders, inventory, purchasing and accounting. Often the data source for automations.
F1 term
- Fine-tuning
- An existing AI model is retrained with your own examples so it matches your tone or terminology. In practice it usually only pays off once RAG is no longer enough.
G1 term
- GDPR
- The European data protection regulation. It defines the legal basis on which personal data may be processed – relevant to every AI rollout.
H1 term
- Hallucination
- A language model invents a plausible-sounding but wrong statement. The remedies are cited sources, clear limits and verification steps in the process.
I1 term
- Integration
- The permanent connection between two systems so data flows automatically – instead of export, rework and manual import.
K1 term
- Knowledge base
- The structured collection of your knowledge: manuals, quotes, policies, FAQs. It is what enables an assistant to answer correctly.
L2 terms
- Large language model
- A language model (LLM) trained on very large amounts of text that understands, summarises and generates text. The workhorse behind most AI applications.
- Low-code
- Development with visual building blocks plus selective code. Fast for standard cases, with limits once logic gets truly specific.
M2 terms
- Machine learning
- A model learns rules from historical data, for example to forecast demand or detect outliers.
- MVP
- Minimum viable product: the smallest version that delivers real value and can be tested in daily use. After that, decisions rest on usage instead of assumptions.
N1 term
- No-code
- Applications and processes are assembled without programming. Ideal for fast prototypes and clearly scoped automations.
O2 terms
- OCR
- Optical character recognition: text from scans, PDFs or photos becomes machine-readable – the basis for automated invoice and document processing.
- Orchestration
- The coordinated interplay of many individual steps, systems and models into one reliable process – including error handling and retries.
P2 terms
- Prompt
- The instruction given to a language model: task, context, format and limits. Good prompts are reproducible and part of the software, not manual work.
- Prompt engineering
- Systematically developing and testing instructions so a model delivers the desired quality reliably and verifiably.
R2 terms
- RAG
- Retrieval augmented generation: the model first searches your own documents and then answers on that basis. This keeps answers current and verifiable.
- ROI
- Return on investment: the ratio of result to effort. For automation it is usually measured in saved hours, shorter lead times or new revenue.
S2 terms
- SaaS
- Software as a service: software is rented and used in the browser instead of self-hosted. Convenient – but a stack of them gets expensive and messy fast.
- Scalability
- A system's ability to stay stable with more users, data or transactions, without having to be rebuilt.
T2 terms
- Token
- The billing and processing unit of language models – roughly a syllable. The length of input and answer determines cost and speed.
- Trigger
- The event that starts an automated process: a new email, a form, a record or a point in time.
V1 term
- Vector database
- Storage for embeddings that finds semantically matching passages in milliseconds. The technical basis of every good internal search.
W2 terms
- Webhook
- An automatic system-to-system notification in real time: as soon as something happens, the next action is triggered – no polling every minute.
- Workflow
- The defined sequence of steps, responsibilities and rules of a business process – the blueprint for every automation.
Missing a term, or is something unclear?
In the intro call we explain every one of them based on your concrete case.