Skip to courses

EXPERRT ACADEMY / INTRODUCTORY COURSE PREVIEW

Learn the foundations of working with always-on AI agents.

Always-on AI agents can continue working on agreed tasks between your conversations with them. This introductory preview is for people who are new to these tools and want to learn how to choose suitable tasks, write clear instructions, check the results and respond when something goes wrong.

Six introductory lessonsNo previous agent experience neededCreate a brief for your own task

You can try the lessons below without enrolling or connecting an AI tool. The full courses are still being developed. Your answers and written work are not saved when you leave or reload this page, so download the workbook if you would like to keep them.

Editorial illustration of a professional working with an AI agent that organises documents, schedules and reports for human review.
Learn how to delegate ongoing work while remaining responsible for the result.

How to use this course preview

Start with the first lesson, or choose a topic from the lesson list. Each lesson explains a principle, shows how it applies in a fictional business and asks you to make a decision. After reading the feedback, use the practical exercise to develop your own instructions in the brief builder further down the page.

Throughout the preview, Northstar Studio is a fictional design business. Its project tracker is a table recording each project’s owner, deadline and status. We use this example to show how an agent could prepare a weekly report while a person remains responsible for checking and sharing it.

Start the first lesson ↓

LESSON 1 OF 6

Choose a task you do every day or week

By the end of this lesson, you should be able to choose a recurring task whose results you can check and whose usefulness you can measure.

A useful responsibility connects a repeated problem to bounded agent work and a result checked by its owner.
A useful responsibility connects a repeated problem to bounded agent work and a result checked by its owner.

When you first work with an always-on agent, it helps to choose a recurring task with a clear result, such as preparing a weekly report from a project tracker. You need to be able to describe the information the agent should use and explain how you will check its work. Before trying the agent, record how long the task takes you and where mistakes usually occur. This gives you a starting point for deciding whether the agent improves the process once you include the time you spend reviewing its results.

Illustration of the fictional Northstar Studio team planning a recurring reporting task with an AI agent.Illustrated case study · Fictional business
A WORKED EXAMPLE

How this applies at Northstar Studio

Northstar Studio is a fictional design business with six people. Each Friday, the owner spends 75 minutes reviewing a project tracker and preparing a report about work that needs attention. The team wants to try an agent that drafts this report, identifies overdue projects and highlights missing owners or deadlines. The owner will continue to check the report, decide priorities and send any client messages. For this example, the team will consider the trial successful if the draft accurately reflects the tracker and takes no more than 20 minutes to prepare and review.

Read more about the reasoning behind this approach

A useful starting task happens often enough to be worth improving, uses information you trust and produces something you can check before anyone relies on it. An ongoing responsibility also needs instructions about when to repeat the work, when to ask for help and when to stop. Different tools handle ongoing work differently: some run on a schedule, while others respond to events or continue pursuing a goal. Before planning a live workflow, check which behaviour your chosen product supports.

Practice question: applying this lesson

You are helping the owner of Northstar Studio choose a first task for an agent. The team has not yet tested how reliably the agent works, so the owner wants a small trial that can be reviewed before it affects clients. Which of the following tasks would you recommend?

Choose the response you would use in this situation. You will receive an explanation of your choice, and you can try another answer afterwards.

APPLY WHAT YOU HAVE LEARNED

Practise with a task of your own

Think of three tasks you regularly repeat, such as gathering updates, preparing a report or organising information. For each task, record how long it takes, where the information comes from and who checks the result. Choose one task you could safely try with an agent, then explain what the agent would prepare and which decisions you would keep with a person.

What to record in your workbook: Keep your three-task comparison and a short explanation of why you selected one task. Include the time it takes today so you can compare it with the agent-assisted process later.

Write your instructions in the brief builder below ↓

WRITE INSTRUCTIONS FOR YOUR AGENT

Create a clear brief for a task you want to delegate

An agent brief is a set of instructions explaining the work you want done, the information the agent should use and the limits it must follow. Use the questions below to draft a brief for your own task, or practise with the fictional Northstar Studio example. You can download your answers with the lesson exercises in a workbook. Use sample information rather than private business records, passwords or access keys.

Describe one recurring task and the result you need. For example, ask it to prepare a weekly project report for you to review.

Name the person responsible for checking the result and deciding what happens when the agent needs help.

Name the files or records it should rely on, how up to date they need to be and what it should do when information is missing.

Describe the output and the checks that establish whether it is correct. For a report, you might require a link to each project record and a list of missing information.

Describe the schedule and timezone, whether your computer must remain on and what should happen if a scheduled run is missed.

Explain what the agent may do itself, what needs your approval and what it must not do. For example, it may draft a message but need approval before sending it.

Explain when the agent should pause, how it should check for an action that may already have completed and who should decide whether to try again.

Record how long the task takes today, the review time and software costs you expect, and the records you will keep to compare results.

You have added notes to 0 of the 8 sections. Use the lesson examples to review whether your instructions are clear enough for someone else to follow.

HOW ASSESSMENT WILL WORK IN THE FULL COURSES

Show how you can apply what you have learned

In the full courses, you will complete a practical project and explain how you checked its results, handled a problem and measured its usefulness. An Experrt reviewer will assess your work against the six skill areas below and ask you to explain a decision or adapt your project to a changed situation. If a skill needs further work, you will receive feedback and an opportunity to improve your submission.

Choosing and describing the taskSetting permissionsMaking the workflow workChecking the resultsHandling failuresMeasuring value and explaining the process

The planned Experrt certificate will be awarded when your practical work meets the requirements in all six areas. It will record the skills assessed by Experrt and will not represent an externally accredited qualification. This introductory preview provides practice only; it does not include assessment or award a certificate.

EXPLORE THE COURSES WE ARE DEVELOPING

Continue learning with a course suited to your work

We are developing separate courses for individual agent platforms and for specific types of work, such as research, content production and small business operations. Each proposed course includes six modules and a final practical project. The planned price is £99 per course; enrolment is not yet available.

PROPOSED COURSE / £99 PLANNED PRICE

Always-On Agents: Foundations

People who want help with tasks they do regularly at work or in their business.

See what the six modules will cover
  1. Choose a task you do every day or week

    What you will produce: A list of possible tasks and a record of how long they take today

  2. Write clear instructions for your agent

    What you will produce: Written instructions and two examples showing what a correct result looks like

  3. Choose what your agent can see and do

    What you will produce: A list of information the agent can use and actions it is allowed to take

  4. Set a schedule and check the results

    What you will produce: A schedule and a checked result showing which records you used

  5. Fix problems and keep costs under control

    What you will produce: A practice problem, a check for repeated work and a cost worksheet

  6. Show how your task works and explain it to someone else

    What you will produce: Your finished project, the records of your checks and a discussion with a reviewer

Before you start: Basic file and browser skills. Practice can use the supplied fictional scenario.

Your final projectSet up a task that prepares a weekly project report. Write its instructions, check the results, practise fixing a problem, track the costs and explain how someone else can run it.

PROPOSED COURSE / £99 PLANNED PRICE

Working with OpenAI Dots

People who want to delegate ongoing research, reporting or project responsibilities in ChatGPT.

See what the six modules will cover
  1. Get to know your dot and where it works

    What you will produce: Access and environment checklist with cloud/local dependencies

  2. Give your dot a task to keep working on

    What you will produce: Dot instructions with checks for a correct result and stop rules

  3. Connect your apps and choose what your dot can do

    What you will produce: Tested app access and Custom Rules record

  4. Check the work and give your dot better information

    What you will produce: Annotated outputs and a corrected follow-up run

  5. Pause your dot and deal with problems

    What you will produce: Missing-source and conflicting-instruction drills

  6. Show how your dot helps with a real task

    What you will produce: Run history, source verification and project explanation

Before you start: Eligible Dots access in a supported region, plus the foundation refresher.

Your final projectConfigure and evaluate a dot that maintains a project briefing, flags meaningful changes and prepares reviewable updates.

PROPOSED COURSE / £99 PLANNED PRICE

Working with Grok Bot

Professionals evaluating ongoing agent work using Grok Bot.

See what the six modules will cover
  1. Check what your bot can do

    What you will produce: A verified record of available features and running requirements

  2. Give your bot a clear task

    What you will produce: Instructions, tasks to leave out and examples of correct results

  3. Choose the information and actions your bot can use

    What you will produce: Source map and integration access record

  4. Check the findings and decide when you need an alert

    What you will produce: Claim-by-claim verification and alert relevance log

  5. Deal with old information and unfinished work

    What you will produce: Practice recovery exercise and intervention record

  6. Compare your bot’s work with how you did the task before

    What you will produce: Final project, cost comparison and individual follow-up

Before you start: Eligible Grok Bot access and verified availability of the integrations used in the course.

Your final projectBuild and evaluate a monitoring and briefing workflow with references that support its findings with clearly scoped actions and a recovery plan.

PROPOSED COURSE / £99 PLANNED PRICE

Working with Meta Muse

People delegating personal planning and administrative work.

See what the six modules will cover
  1. Set up Muse and choose something you need help with

    What you will produce: A clearly scoped goal and verified access checklist

  2. Tell Muse what it needs to know and remember

    What you will produce: List of useful information and a demonstrated correction or forgetting action

  3. Choose what Muse can do and when it must ask you

    What you will produce: Personal app permissions and action decision table

  4. Make a plan that fits your week

    What you will produce: A checked plan based on fictional constraints

  5. Update the plan when your needs change

    What you will produce: Changed-goal and conflicting-calendar drills

  6. Show how Muse helps you stay organised

    What you will produce: Plan comparison, review record and project explanation

Before you start: Eligible Muse access; fictional calendar and message examples are supplied.

Your final projectCreate a personal planning workflow that proposes a weekly plan, resolves conflicts and leaves actions that affect other people or records for approval.

PROPOSED COURSE / £99 PLANNED PRICE

Working with Claude Cowork

Office professionals automating recurring file and reporting work.

See what the six modules will cover
  1. Set up your files and check how Cowork runs

    What you will produce: Working folder, access boundaries and verified local or remote execution requirements

  2. Ask Cowork to help with a task at work

    What you will produce: A task brief and accepted sample result

  3. Save useful instructions and connect the tools you need

    What you will produce: A reusable skill or saved instructions and tool test

  4. Set up a regular task and check the result

    What you will produce: Schedule evidence and a source-checked report

  5. Fix missing files, failed tasks and incorrect results

    What you will produce: A failed-run recovery record and safe revision

  6. Show how Cowork prepares a report each week

    What you will produce: Final project, handover and individual follow-up

Before you start: Eligible Cowork access and an environment that meets the documented execution requirements.

Your final projectProduce a recurring weekly briefing from supplied files, with traceable facts, a reusable saved instructions and missed-run recovery.

PROPOSED COURSE / £99 PLANNED PRICE

Business Agents with Microsoft Copilot Studio

Business process owners and Microsoft ecosystem makers.

See what the six modules will cover
  1. Describe how the task works today

    What you will produce: Request handling map, ownership and success measures

  2. Build your agent and give it the right information

    What you will produce: Configured agent and example answers linked to their sources

  3. Connect tools and choose which actions need approval

    What you will produce: Connector tests and action permissions

  4. Choose when the task starts and who checks it

    What you will produce: End-to-end fictional request run

  5. Test what happens when something goes wrong

    What you will produce: Invalid-input, duplicate and failure tests

  6. Plan a small trial with your team

    What you will produce: Final project solution, reversal and recovery plan and owner handover

Before you start: A suitable development environment, licensing and basic process mapping skills.

Your final projectBuild a fictional service request handling agent that validates a request, prepares routing and escalates exceptions.

PROPOSED COURSE / £99 PLANNED PRICE

Running Your Own Agent with OpenClaw

Technical learners who want to operate and maintain their own agent.

See what the six modules will cover
  1. Choose a computer and set up your agent

    What you will produce: Environment inventory and successful setup checks

  2. Connect an AI model and choose the tools it can use

    What you will produce: Configuration record with private details removed, plus access checks

  3. Give your agent instructions and useful information

    What you will produce: Saved instructions and memory correction example

  4. Set up regular checks and useful alerts

    What you will produce: Test runs and useful-notification criteria

  5. Track costs and get your agent working again

    What you will produce: Interrupted-run, budget and duplicate tests

  6. Keep your agent running and explain it to the next person

    What you will produce: Final project, maintenance guide and reversal and recovery demonstration

Before you start: Command-line confidence, a supported host and a model provider account with a controlled budget.

Your final projectOperate a clearly scoped monitoring agent with a regular check-in (heartbeat), explicit tool permissions, run records and a tested recovery procedure.

PROPOSED COURSE / £99 PLANNED PRICE

Managing Agents Reliably

Anyone responsible for agents that continue working between conversations.

See what the six modules will cover
  1. Decide who is responsible and write down the task

    What you will produce: Responsibility inventory and accountable owner map

  2. Choose what your agent can see and do

    What you will produce: Table of allowed actions and boundary challenge results

  3. Make a set of examples to test your agent

    What you will produce: Success, ambiguity and problem-situation set of checks

  4. Fix a failed task without doing the same work twice

    What you will produce: Incident timeline and a tested recovery procedure

  5. Track spending, checking time and useful alerts

    What you will produce: Cost model and notification tuning comparison

  6. Check that the task works and explain it to someone else

    What you will produce: Audit findings, corrected workflow and project explanation

Before you start: One configured agent or the supplied simulation and fictional run logs.

Your final projectAudit and improve a flawed workflow, then demonstrate permission checks, verification, incident recovery and measurable value.

PROPOSED COURSE / £99 PLANNED PRICE

Agents for Small Business Operations

Small business owners and operations staff.

See what the six modules will cover
  1. Find a task that takes up too much time

    What you will produce: Task starting measurements and ranked opportunities

  2. Get your business records ready

    What you will produce: Field definitions and source ownership map

  3. Decide who receives each task and who checks the work

    What you will produce: Workflow and permission decision table

  4. Prepare a useful report each week

    What you will produce: Checkd report with traceable claims

  5. Deal with repeated tasks and missing information

    What you will produce: Exception log and recovery demonstration

  6. Check whether the agent helps and plan a small trial

    What you will produce: Final project pack and a 30-day adoption plan

Before you start: Basic spreadsheet skills; use fictional business records for practice.

Your final projectBuild a weekly operations pack that identifies outstanding work and prepares owner updates without silently changing business records.

PROPOSED COURSE / £99 PLANNED PRICE

Research and Market Monitoring with Agents

Researchers, business owners and market analysts.

See what the six modules will cover
  1. Choose a research question and decide what counts as a reliable answer

    What you will produce: Research contract and relevance criteria

  2. Choose your sources and check when they were updated

    What you will produce: List of sources with publication and event dates

  3. Set up regular checks for important changes

    What you will produce: Starting measurements snapshot and change log

  4. Check the findings and explain what is still unclear

    What you will produce: Claim-to-source table and challenged finding

  5. Make alerts useful and deal with missing sources

    What you will produce: Alert usefulness log and fallback test

  6. Show how you track changes and check the findings

    What you will produce: Final project brief, test set and individual follow-up

Before you start: Basic source evaluation skills; no live commercial dataset is required.

Your final projectCreate a monitoring brief that distinguishes new evidence, repeated reports and uncertain claims, and sends only meaningful alerts.

PROPOSED COURSE / £99 PLANNED PRICE

Content Operations with Agents

Content creators and marketing teams.

See what the six modules will cover
  1. Decide who you are writing for and how you want to sound

    What you will produce: Brand brief and editorial examples of correct results

  2. Check the facts before you write

    What you will produce: Source inventory and fact-check table

  3. Write your drafts and choose helpful images

    What you will produce: Draft set, image brief and rights/source record

  4. Decide who checks and approves each draft

    What you will produce: Editorial workflow with publishing boundary

  5. Correct the content and avoid publishing it twice

    What you will produce: Revision history and duplicate test

  6. Finish your content and check how much time it took

    What you will produce: Final project pack and quality/effort comparison

Before you start: A fictional brand brief is supplied; basic editing skills are sufficient.

Your final projectBuild a repeatable content preparation workflow with source checks, editorial review, accessible assets and revision tracking.

PROPOSED COURSE / £99 PLANNED PRICE

Agents for Developers

Developers delegating clearly scoped software work and maintenance.

See what the six modules will cover
  1. Describe the code change and how you will test it

    What you will produce: Issue brief and meaningful test plan

  2. Set up a separate place to work on the code

    What you will produce: Workspace, secret boundaries and reversal and recovery check

  3. Ask your agent to make a change you can review

    What you will produce: Bounded change and reasoning record

  4. Check the code, tests and any added packages

    What you will produce: Diff review and verified test results

  5. Fix failed checks and avoid repeating the same change

    What you will produce: Failure recovery and duplicate change prevention

  6. Explain your change and show that it works

    What you will produce: Final project pull request, handover and individual follow-up

Before you start: Git, code review and testing experience; a disposable practice repository is required.

Your final projectDelegate a maintenance change, inspect its diff and tests, and demonstrate recovery without granting unattended production deployment.

PROPOSED COURSE / £99 PLANNED PRICE

Coordinating Multiple Agents

Experienced agent users coordinating research, drafting and review responsibilities.

See what the six modules will cover
  1. Decide whether you need more than one agent

    What you will produce: Single-agent starting measurements and justified role split

  2. Give each agent a clear job and the information it needs

    What you will produce: Instructions for each agent and list of records to follow

  3. Decide how agents pass work to each other

    What you will produce: Format for passing work between agents and rejected-input example

  4. Run the agents together and check their work

    What you will produce: End-to-end run with traceable responsibility

  5. Deal with disagreements and work that gets stuck

    What you will produce: Conflict and limits on repeated attempts drills with cost records

  6. Compare the team with using one agent

    What you will produce: Final project, comparative evaluation and owner handover

Before you start: Completion of one practical agent project or equivalent experience.

Your final projectCoordinate a researcher, drafter and reviewer through explicit handoffs, clearly scoped authority and an accountable final owner.