AI agents and assistants by profession

Library of agents and assistants

AI agents and assistants by profession: 67 ready-to-configure examples

AI assistants and agents designed for real-world work in large organizations, categorized by profession: HR, finance, sales, legal, management, project management, purchasing, and many more. Each example links to a free guide with instructions for gluing, source code connections, safeguards, and implementation steps in Copilot, ChatGPT, Claude, and Mistral.

  • 30 assistants
  • 37 connected agents
  • 21
  • Free, without registration

Assistant or agent: what's the difference?

Both are based on the same idea: to write down once and for all what the AI ​​needs to know and do, instead of having to explain everything again in every conversation. They don't go to the same place.

  • The assistant It is configured with instructions and your reference documents (procedures, playbook, repository). It responds when queried, always with the same precision. It can be created in minutes, without any special privileges: Agent Builder in Copilot, a ChatGPT project, or Claude, a Vibe agent.
  • The connected agent It is connected to the organization's tools. It meets at least one of these three criteria: it reads from a changing source (calendar, CRM, register), it interacts with a tool (creates a task, prepares a draft), or it runs autonomously at regular intervals. It generally requires Copilot Studio and approval from the IT department.

Practical rule: start with the assistant. When you find yourself copying the same data to them every week, it's time to move on to the agent. For the one-off task, a prompt from the library That's often enough.

The 67 assistants and agents, job by job

Each card opens the complete sheet: the role, the instructions to paste, the sources to connect, the limits to set and an example of the result.

Before deploying an agent

The guidelines apply four safeguards that are best put in place from the outset:

  • Limited access : the agent only reads the sources that the task requires, with the rights of the person using it.
  • Human validation before any irreversible action: the agent prepares a draft, a person sends, publishes or deletes.
  • A test perimeter : one team, one project, a few weeks, before expanding.
  • An owner named who monitors the results, corrects the instructions and decides to stop if necessary.

For projects that involve multiple documents and produce multiple deliverables, see also the examples of missions Claude Cowork by job. The complete library gathers prompts, assistants, agents and Cowork, classified in the same way.

Frequently Asked Questions about AI Agents and Assistants

What is the difference between an AI assistant and an AI agent?

An assistant is configured with instructions and reference documents and responds when queried. An agent is connected to the organization's tools: it reads a changing source, acts within a tool, or executes on its own at regular intervals.

What tools can be used to create these assistants and agents?

In Microsoft 365, Agent Builder is used for assistants, and Copilot Studio for connected agents. These tools can also be implemented in a ChatGPT or Claude project and in a Mistral Vibe agent. Each guide details the steps for each tool.

Do you need to know how to code to create an AI agent?

No. An assistant is created by pasting instructions and adding documents. A connected agent primarily requires access rights to the source code and, in most large organizations, approval from the IT department.

Are these fact sheets free?

Yes, all the library's records are freely accessible, without registration and without usage limits.

What precautions should be taken before deploying an agent?

Limit access to what is strictly necessary, maintain human validation before any irreversible action, test it on a restricted perimeter and appoint an owner responsible for monitoring it.

Build your first agents with your teams

The Agent Party enables your teams to build their own assistants and agents, based on their real-life cases, within your company's tool.

Back to top
The entire NextStart.AI libraryEight consulting missions, seven pillars, around twenty formats, fifteen professions, five tools
Nine detailed customer case studiesLarge groups, public agencies and networks: from acculturation to production

Insurance and social protection

Health and life sciences

Public sector and research

The NextStart.AI collectiveConsultants, trainers and industry experts: more than 4,850 professionals trained since 2023
All free resourcesDiagnostics, Finders, prompt library, toolkits, kits for specific roles and professions, and for managing an AI program
The NextStart.AI blogComparisons of tools, practices, agents, governance: analyses and feedback from the field
Let's talk about your AI projectResponse within 24 business hours, written quote after initial discussion
AI agents in businessUnderstanding, choosing the tool, training the teams, deploying
Delegate entire tasksClaude Cowork, Copilot Cowork, ChatGPT Work, Vibe Work