Glossary
AI in recruitment, in plain words.
Every vendor defines these terms to suit what they sell. These definitions are written for the person who has to make the thing work on Monday.
Agentic AI
Agentic AI is the category of AI systems that plan and execute sequences of actions towards a goal instead of producing one response. In recruitment it covers anything that reads your files, calls other tools, and writes results back. The term describes a capability level, not a product.
AI agent
An AI agent is a model that carries out a multi-step task on its own rather than answering a single question. You describe the outcome, it decides the steps, uses tools, and reports back. The difference from a chatbot is that an agent acts, and acting means it can be wrong in ways that leave traces.
AI literacy
AI literacy is the ability of the people operating an AI system to explain what it does, where its data came from, how it fails, and who made the decision. Under the EU AI Act it is a requirement for organisations deploying AI, not a nice-to-have. It is measured by what a team can demonstrate, not by certificates collected.
AI Recruitment Engineer
An AI Recruitment Engineer is a recruiter who builds systems instead of doing tasks. Rather than prompting an AI tool one request at a time, they connect their ATS, sourcing tools and outreach into automations that keep running without them. It is a working method, not a job title, and any recruiter can learn it without writing code.
AI sourcing
AI sourcing is finding candidates by describing the role in plain language and letting a model surface semantically similar profiles. It finds people whose vocabulary you would not have guessed. It misses people who wrote very little about themselves, which is the same group Boolean already missed.
Algorithmic bias in hiring
Algorithmic bias in hiring is the tendency of a model to reproduce patterns in the data it learned from, including the ones you would not defend out loud. It rarely looks like discrimination in the output. It looks like a shortlist that quietly resembles the people you already hired.
ATS (Applicant Tracking System)
An ATS is the database and workflow tool that holds your candidates, vacancies and hiring stages. It is the system of record for recruitment, which makes it the source of most AI outputs and most AI errors. Its data quality sets the ceiling on anything you automate on top of it.
Blind sourcing
Blind sourcing is hiding candidate photos and names during the search and review stage to reduce first-impression bias. LinkedIn Recruiter supports it as a setting that an admin can enforce for a whole organisation. It removes one bias from one step and leaves the rest of the funnel untouched.
Boolean search
Boolean search finds candidates by combining keywords with operators such as AND, OR and NOT. It is precise, auditable and entirely dependent on you guessing the words a candidate used about themselves. When it returns nothing you can read the string and see why, which is more than most AI search offers.
Claude Cowork
Claude Cowork is the mode in Claude where the assistant works as an agent on a folder of your real files instead of answering in a chat window. You describe the outcome, it works through the steps and saves finished files back. It requires a paid Claude plan, and the desktop app is where it reaches local files.
Connector (MCP)
A connector is a defined link that lets an AI model reach a system such as your ATS, calendar or drive. Model Context Protocol, or MCP, is the open standard those connectors are built on. Connector and MCP server refer to the same thing in practice.
Data hygiene
Data hygiene is the practice of keeping candidate and vacancy records accurate, consistent and current. In recruitment it is the difference between an AI that helps and one that confidently repeats a mistake from four years ago. AI does not fix old data. It scales the damage.
EU AI Act
The EU AI Act is the European regulation that classifies AI systems by risk and attaches obligations to each class. AI used for recruitment and candidate selection sits in the high-risk category, which brings requirements around oversight, transparency and record keeping. Its deadlines have shifted more than once, so plan around the capability rather than the date.
GDPR in recruitment
GDPR governs how candidate personal data may be collected, stored and processed, and it applies in full to anything you paste into an AI tool. Candidates keep rights over their data regardless of which system holds it, including the right to an explanation of decisions that affect them. Retention limits apply to your talent pool as much as to your ATS.
Generative Engine Optimization (GEO)
GEO is the practice of making content that AI answer engines will quote, rather than content that only ranks in a list of blue links. It rewards clear definitions, answer-first passages and structured data over keyword density. For recruitment brands it matters because candidates and hiring managers increasingly ask a model before they ask a search engine.
Hallucination
A hallucination is a confident, fluent output that is factually wrong. Models produce them because they predict plausible text rather than retrieve verified facts. In recruitment the dangerous ones are invented employment dates, misattributed quotes and fabricated company details, because they read exactly like the true ones.
High-risk AI system
A high-risk AI system is one the EU AI Act places in its most regulated tier because it affects people's access to things like employment, credit or education. Recruitment and candidate selection are named explicitly. The classification follows the use, not the technology, so an ordinary language model becomes high-risk the moment you point it at a shortlist.
Human in the loop
Human in the loop means a person reviews an AI output, has the authority to overrule it, and leaves a record of what they decided. All three parts are required. A workflow where someone clicks approve without the ability or the information to disagree is a rubber stamp, not oversight.
n8n
n8n is a workflow automation tool that connects apps and AI models through a visual editor rather than code. Recruiters use it to build the automations that sit between their ATS, their sourcing tools and their outreach. It is the tool used in the Klikwork bootcamps because you can read what you built six months later.
Prompt engineering
Prompt engineering is the practice of writing instructions that get a reliable result from a language model. In recruitment its value is overstated: a good prompt run by hand every week is still manual work. The durable version of a good prompt is a skill or an automation.
Quality of hire
Quality of hire measures how well the people you hired actually perform and stay. It is the only recruitment metric that tracks the outcome rather than the process, and the hardest one to collect honestly. Teams that skip it end up optimising a funnel with no idea whether it produces good hires.
Screening
Screening is assessing applicants against role criteria to decide who moves forward. It is the step where AI is most tempting and most regulated, because a screening decision is a decision about a person. Use AI to structure and prepare the information, and keep a human making the call.
Semantic search
Semantic search finds results by meaning rather than by matching exact words. It surfaces the candidate who calls themselves a talent partner when you searched for recruiter. Its weakness is the mirror of its strength: it rewards profiles written in the same style as the data it learned from.
Setup Scan
A Setup Scan is a diagnostic review of a recruitment team's AI tools, workflows and data before anything new is added. It establishes what is actually broken, which is usually different from what the team assumed. It is the step most teams skip, which is why their tools change and their results do not.
Skill (Claude)
A skill is a reusable instruction set that teaches an AI assistant how you want a specific task done. Once written, it runs the same way every time without you re-explaining. For a recruiter it turns a prompt you keep rewriting into something the team can share.
Talent pool
A talent pool is a maintained group of candidates you have already engaged and expect to approach again. Its value comes entirely from being current, which is why most pools are a liability rather than an asset. An outdated pool trains your AI on people who moved on years ago.
Time to hire
Time to hire measures the days between a candidate entering your process and accepting an offer. It is the metric most often quoted to justify AI spending and the one most easily improved by cutting the wrong step. Read it next to quality of hire or it will reward speed over outcome.
Vibe coding
Vibe coding is building working software by describing what you want to an AI model instead of writing the code yourself. Recruiters use it to make small internal tools, dashboards and scrapers without a developer. It works well for things you can throw away and badly for things other people depend on.
Workflow automation
Workflow automation is connecting the steps of a recurring task so it runs without a person triggering each one. In recruitment that usually means intake preparation, research, reporting or follow-up. It is different from AI: the automation carries the work, the AI handles the judgement inside a step.
X-Ray search
X-Ray search uses a general search engine to look inside a specific site, most often LinkedIn, using a site: operator. It reaches profiles without needing a paid seat on the platform itself. Results depend on what the search engine has indexed, so it complements platform search rather than replacing it.