Agents, tools, and orchestration
A chatbot answers questions. An agent goes and does things. That is where most of the value in modern AI sits, and it is also where the risk sits, because a system that acts on its own multiplies small errors into large ones. Building agents well turns out to be a discipline of holding them back.
Say whether a proposed agent is safe to build, and name the specific constraints it needs before it ships.
Lessons 1 and 2. You will want to be comfortable with retrieval before adding autonomy on top of it.
Plain-English mode
Each section opens with a short summary before the detail. Turn this off once the vocabulary feels familiar.
An agent is a model put in a loop, given the ability to use tools, and left to keep going until a job is done. A chatbot answers. An agent acts.
An ordinary model call is a closed loop: text goes in, text comes out, nothing happens in the world. An is that same model placed inside a loop where it can take actions and see what came back, repeating until the job is done. In practice that means the model can decide to use a tool: search a database, call another system, send an email, run some code, look up a document. It reads the result, then decides what to do next.
The distinction in one line: a chatbot answers, an agent gets things done. "Book the appointment." "Pull this applicant's records, check them against the regulation, and draft a decision." "Research this law across three jurisdictions and tell me where the gaps are." None of those are answerable by talking. They require doing.
The basic loop is worth being able to draw on a whiteboard, because it demystifies the whole category. Decide on an action, meaning which tool and what to pass it. Take the action. Read the result. Repeat until the goal is met or a stop condition trips. Then write the final answer. That is genuinely all an agent is. Everything else in this lesson is about controlling that loop.
The rest of the lesson covers how tools actually work, the patterns for organising agents, how memory gets built, the emerging standard for plugging tools in, and the reliability problem that decides whether any of this is safe to ship. Those sections are part of the paid course.
Answer in your head first, then reveal. Getting it wrong is the part that makes it stick.
What single capability turns a chatbot into an agent?
The rest of this lesson
- How an agent actually does anything
- The three ways agents think
- Using several agents together
- Memory is something you build
- MCP: the standard plug
- The work that does not happen in three seconds
- Why agents are hard: errors compound
- One worked example: helping planning officers
Unlock “Agents, tools, and orchestration” and the whole playbook
One payment unlocks all 12 AI lessons in full, plus every career map, 90-day plan, and template in the Clinical Pivot Playbook.
This is educational career guidance, not a promise of a job, a salary, or any specific outcome, and not financial or legal advice. Your results depend on your own effort and circumstances. Make big decisions with proper professional advice where it matters.
Teaching a model your stuff: prompting, RAG, and fine-tuning
Evaluation: knowing whether it actually works
Unlock every AI lesson
One payment unlocks all 12 AI lessons in full, plus the entire Clinical Pivot Playbook: 8 career maps, 8 personalised 90-day plans, CV and LinkedIn templates, the cold-message bank, the interview answers, the glossary, and the startup list.