AI Recruiting Academy › Module 1: AI basics for recruiters
The short answer
A chatbot answers when you ask it something. A copilot sits inside a tool you already use and drafts or suggests while you work. An AI agent is handed a goal and works through several steps on its own, using other tools along the way.
The further along that line you go, the more work gets done without you, and the more important it is to decide where the system must stop and wait for a human.
Why do these three words get mixed up?
Because vendors use them loosely. A product called an “agent” in a demo may be a chatbot with a nicer screen, and a real agent may be sold as a “copilot”. The labels matter less than two questions: what can this system do without me, and where does it stop?
All three usually run on the same kind of engine, a large language model. The difference is what is built around the engine: how much it is allowed to do, which tools it can reach and how often it checks back with a person.
What is a chatbot?
A chatbot is a conversation. You type, it replies, and nothing happens outside the chat window unless you copy the answer somewhere yourself.
Recruiters meet two kinds. The first is the general assistant, such as ChatGPT, Claude, Gemini or Grok, which you use to draft a job description or rework an outreach message. The second is the candidate facing chatbot on a careers page that answers questions about benefits, collects basic details and books a screening call.
A chatbot is the lowest risk place to start, because everything it produces passes through your hands before anyone else sees it.
What is a copilot?
A copilot is AI built into software you already work in. Your ATS offers to summarize a candidate. Your email tool suggests a reply. Your sourcing platform drafts a first message from the profile you are viewing.
The copilot can see the context in front of you, which makes its suggestions more useful than a blank chatbot, but you still take every action. You click send. You move the candidate. You decide.
Most recruiting software added copilot features over the last two years, so many recruiters already use AI this way without calling it that.
What is an AI agent?
An agent gets a goal instead of a single instruction, and it decides the steps. Anthropic, which builds the Claude models, draws the line clearly in its guide to building effective agents: in a workflow, the steps are fixed in advance by the people who built it, while an agent directs its own process and chooses which tools to use as it goes.
A recruiting example makes it concrete. You tell an agent: “Find 30 senior backend engineers in Austin who have worked at a startup, find their work emails, draft a first message for each and put them in a sequence for my approval.” The agent searches, reads profiles, checks results against your brief, calls an email finder, writes drafts and loads them into your outreach tool. Then it stops and waits for you.
That last step is the part that separates a well built agent from a risky one. We cover real agent workflows by stage in our guide to AI agents for recruiting, and Module 5 goes deeper.
How do chatbots, copilots and agents compare?
| Chatbot | Copilot | AI agent | |
|---|---|---|---|
| How it starts | You ask a question | You work; it offers help | You give it a goal |
| What it can touch | Only the chat | The tool it lives in | Several tools and data sources |
| Who takes each action | You | You | The agent, until a checkpoint |
| Recruiting example | Rewrite this job ad | Summarize this candidate in my ATS | Build, enrich and draft a shortlist |
| Main risk | Wrong facts you copy | Suggestions accepted without reading | Actions taken at scale before anyone checks |
| Best first use | Drafting and editing | Summaries and replies | Research and list building with approval |
Is an agent the same as automation?
No, and the difference matters when you buy. Traditional automation, the kind you build in a workflow tool, follows fixed rules: when a candidate reaches the offer stage, create a task and send a template. It does exactly the same thing every time, which makes it predictable and easy to audit.
An agent can handle messier work, like judging whether a profile really matches a brief, but it may take a different path each time. Anthropic’s advice in the same guide is to use the simplest approach that works and only add agent style freedom when the task needs it. For recruiters that means a plain rule is often the better choice for anything repetitive and predictable. See the workflow automation category for rule based tools.
Where should an AI agent stop and wait for a human?
Build a checkpoint anywhere an action is hard to undo, reaches a real person or affects a hiring decision.
| Action | Checkpoint? | Why |
|---|---|---|
| Searching and reading profiles | No | Nothing leaves your desk |
| Drafting messages | No | Drafts are easy to fix |
| Sending messages to candidates or clients | Yes | A bad message cannot be unsent and damages your name |
| Rejecting or advancing a candidate | Yes | It is a hiring decision with legal weight |
| Changing records in your ATS | Yes, at first | Bad data spreads to every search after it |
| Spending credits or money | Yes | Agents can burn a month of credits in an afternoon |
Which one should you start with?
Start with a chatbot for drafting, turn on the copilot features in the tools you already pay for, and try an agent only for research and list building where it has to show you its work before anything goes out. Most recruiters get the biggest early win from the copilot features they already have and have never switched on.
Quick answers
Is Grok Bot a chatbot or an agent?
Grok Bot is an agent. It runs on its own cloud computer and can work across apps, unlike the Grok chat assistant. Our explainer on Grok Bot vs a chatbot covers the difference.
Is an AI interviewer an agent?
Usually it sits in between. It holds a conversation on its own, but inside a fixed script and scoring guide set by the employer. Module 3 covers AI interviewing.
Do agents need special connections to my tools?
Yes. An agent needs a way into your ATS, inbox and data sources, usually through an API or MCP. Lesson 5 explains both.
Will AI agents replace sourcers?
They are taking over the list building part of sourcing. Judging fit, reading between the lines of a profile and winning over a passive candidate still sit with people.