- 1: Part 1, Understanding AI agentsPart 1, Understanding AI agents
- 2: Part 2, The six principles
- 3: Part 3, Introducing an agent
- A–E: Appendices
Part 1, Understanding AI agents
Part 3, Introducing an agent
Chapter 1.1
What an agent is
Scope and limitations of this guide
This guide provides general information on organising work with AI agents, not advice tailored to a particular organisation, system or decision. It is not legal or regulatory advice.
Its examples, checklists and templates do not establish that a proposed use is safe, lawful or suitable. You need to assess and test the proposed use in your own circumstances. Following the guide does not by itself meet the requirements of any law or standard.
The guide is provided in good faith, using information available at the time of publication or revision. AI systems, provider terms and legal requirements can change. Before acting, check the current requirements that apply to your organisation and country, confirm what the system can do and test the proposed use. You are responsible for how you apply the guide and should obtain professional advice where appropriate.
Reading, downloading or using the guide does not by itself create an adviser–client relationship with Tom Clowry or Considered Ideas Ltd.
To the fullest extent permitted by law, Tom Clowry and Considered Ideas Ltd accept no liability for loss or damage arising from use of, or reliance on, this guide. Nothing in this notice excludes or limits liability for death or personal injury caused by negligence, fraud or fraudulent misrepresentation, or any other liability that cannot lawfully be excluded or limited.
In this chapter
This chapter explains what an agent is, what agents make possible, the parts of an agent and the model it uses.
The six principles draw on these ideas, and each principle refers back to the ones it depends on.
Assistants, automated processes and agents#
An AI agent uses an AI model and software tools to carry out a task over several steps. What distinguishes it from other uses of AI is who decides the steps.
An AI assistant
Answers a question or produces a draft in response to each request.
An automated process
Follows steps and decision rules set in advance.
An AI agent
Uses the information available and the result of each step to determine the next.
An agent can therefore handle work that varies from case to case. You cannot always predict each step it will take, which is why its instructions, permissions and checks matter.
What agents make possible#
Agents can undertake work that has until now depended on human flexibility. Given suitable instructions and permissions, they can:
Interpret information
Read enquiries, records and documents, and work out what each requires.
Transform and create outputs
Turn notes into a report, figures into a comparison or a request into a draft reply.
Apply judgement within defined boundaries
Assess cases against agreed criteria, make decisions within their authority and refer cases outside that authority for review.
Coordinate actions
Carry a task across several systems, such as checking an order, arranging a delivery and updating the record.
Saving time is one benefit. Others include examining more cases, using information while a decision is still open and offering explanations suited to different needs.
The parts of an agent#
An agent combines an AI model with the instructions, information, tools, permissions and checks around it, known collectively as the harness. Together, these determine what information it can use, which actions it can take and when it must stop.
The model#
The model is the part that reads, reasons and writes. It has been trained on very large amounts of text, and often images and sound, to predict what is likely to come next. That training allows it to summarise, draft, answer questions, work through problems and choose actions.
The other parts of an agent supply current information, memory and the ability to act. A model can reason about information and compare claims with evidence supplied to it. Its response can still contain errors, so a fluent answer is not sufficient evidence that a claim is correct.
Model capabilities and trade-offs#
AI services offer a choice of models, and one agent can use different models for different steps. Models differ in capability, speed, cost and the kinds of information they can process. A model can combine several of the capabilities below.
Models designed for a wide range of tasks, including analysis, writing and coding. Their performance, speed and cost vary.
Smaller models can offer lower cost and faster responses. They are useful for well-defined or high-volume tasks where they meet the required standard.
Some models can work through a problem in greater depth before responding. This can improve performance on difficult tasks, but can increase response time and cost.
Some models can process images, audio or video alongside text. The formats supported vary by model and service.
Test a model on examples of your own work before relying on it, because general comparisons may not reflect your tasks.
Where a model runs#
A model runs either on a provider’s systems or on your own. The choice affects who processes your information, who maintains the system and which models you can use.
A provider runs the model.
Your organisation runs the model.
A model’s weights are its trained parameters. Some developers publish them under a licence, as open-weight models. Whichever way a model is run, connected tools can still send information elsewhere.
© 2026 Considered Ideas Ltd. Licensed under the Creative Commons Attribution 4.0 International licence (CC BY 4.0). This guide provides general information and is not legal or regulatory advice. The guide’s home page gives the suggested credit.
Scope and limitations of this guide
This guide provides general information on organising work with AI agents, not advice tailored to a particular organisation, system or decision. It is not legal or regulatory advice.
Its examples, checklists and templates do not establish that a proposed use is safe, lawful or suitable. You need to assess and test the proposed use in your own circumstances. Following the guide does not by itself meet the requirements of any law or standard.
The guide is provided in good faith, using information available at the time of publication or revision. AI systems, provider terms and legal requirements can change. Before acting, check the current requirements that apply to your organisation and country, confirm what the system can do and test the proposed use. You are responsible for how you apply the guide and should obtain professional advice where appropriate.
Reading, downloading or using the guide does not by itself create an adviser–client relationship with Tom Clowry or Considered Ideas Ltd.
To the fullest extent permitted by law, Tom Clowry and Considered Ideas Ltd accept no liability for loss or damage arising from use of, or reliance on, this guide. Nothing in this notice excludes or limits liability for death or personal injury caused by negligence, fraud or fraudulent misrepresentation, or any other liability that cannot lawfully be excluded or limited.

