01
Artificial Intelligence (AI)
Software that can perform tasks that normally require human intelligence — reading, writing, reasoning, recognising images, or making decisions.
e.g. A system that reads a site diary and summarises the key issues.
02
Large Language Model (LLM)
The AI engine behind chatbots like Claude and ChatGPT. Trained on vast amounts of text to understand and generate human language.
e.g. Claude is an LLM made by Anthropic.
03
Prompt
The instruction or question you give an AI model. The quality of the prompt directly determines the quality of the output.
e.g. "Summarise this delay event in 3 bullet points for the client."
04
Hallucination
When an AI confidently produces incorrect or made-up information. Always verify AI outputs against source data for anything critical.
e.g. An AI citing a regulation that doesn't exist.
05
Machine Learning (ML)
A subset of AI where a system learns patterns from data rather than being explicitly programmed. The more data, the better it gets.
e.g. A delay predictor trained on 10 years of project data.
06
AI Agent
An AI that doesn't just respond — it takes actions. It can browse the web, read files, send emails, and complete multi-step tasks autonomously.
e.g. An agent that pulls P6 data and emails a report every Monday.
07
Automation
Setting up a system to complete a task automatically, without human input each time. AI makes automation smarter and more flexible.
e.g. Auto-generating a weekly progress report from raw data.
08
Workflow
A defined sequence of steps to complete a task. In AI automation, workflows connect triggers, logic, and actions into a repeatable process.
e.g. Form submission → AI summary → Slack notification.
09
Context Window
The amount of text an AI can "read" and remember in a single session. Larger context = the AI can process longer documents at once.
e.g. Claude can read a 200-page specification in one go.
10
MCP (Model Context Protocol)
A standard that lets AI models connect to external tools and data sources — like giving the AI hands to reach into your systems.
e.g. Claude accessing your Google Drive files via an MCP connector.
11
Training Data
The information an AI model was "taught" on. The training data shapes what the model knows, how it writes, and what biases it may carry.
e.g. An LLM trained on internet text knows general knowledge but not your project data.
12
RAG (Retrieval-Augmented Generation)
Giving an AI access to your own documents so it can answer questions based on your specific data, not just its training knowledge.
e.g. Asking Claude questions about your contract and getting accurate answers.
13
Webhook
A URL that receives data when something happens in another app — the digital equivalent of a notification that triggers an action.
e.g. A new Airtable record → webhook → n8n workflow fires.
14
Token
The unit AI models use to measure text. Roughly 4 characters or ¾ of a word. Pricing and context limits are measured in tokens.
e.g. "The project was delayed" = 6 tokens.
15
Generative AI
AI that creates new content — text, images, code, or audio — rather than just analysing existing data.
e.g. Claude writing a delay notice from your bullet points.
16
Tool Use / Function Calling
The ability for an AI to call external tools mid-conversation — run a calculation, search the web, or query a database — and use the result.
e.g. Claude running a Python script to analyse schedule float.
17
API (Application Programming Interface)
A way for two software systems to talk to each other. Most SaaS platforms have APIs that n8n and AI tools use to exchange data.
e.g. n8n calling the OpenAI API to summarise a document.
18
System Prompt
Hidden instructions given to an AI before the conversation starts — sets the AI's role, tone, and rules for a specific use case.
e.g. "You are a construction planning assistant. Only discuss schedule-related topics."
19
Autonomous Agent
An AI that can plan and execute a complex, multi-step task on its own — deciding what to do next based on results, without human input at each step.
e.g. An agent that checks for schedule delays, identifies root causes, and drafts a report.
20
Fine-tuning
Training an existing AI model on your specific data so it becomes specialised for your domain — speaks your language, knows your formats.
e.g. A model fine-tuned on construction specifications and contract language.