Overview
Claude is more than a chat window. Projects, Artifacts, Skills, connectors and Cowork can take on a great deal of recurring work in administration, finance, purchasing and HR. This course is a guided map of those building blocks: what each one is for, how they fit together, and how to choose the right layer for a given task.
Day 1 covers the foundations for daily use: how language models behave, good instructions, Projects as shared memory and Artifacts for analyses and reports. Day 2 extends Claude with your first custom Skill, connectors to mailbox, calendar and files, Cowork and routines, and the team rules that keep it all safe. You leave with a working Project, a custom Skill, a routine and a plan for the next 30 days.
Who it's for
People in business-support roles who use Claude, or are introducing it in their company, and want to go beyond the chat. It is especially suited to:
- Administration and office-management staff
- Finance and purchasing teams
- HR professionals
- Project and process managers
Prerequisites
- No programming skills are needed
- A real use case from your daily work to bring along
- A Claude account on a paid plan (see What you'll need)
By the end of this course you will be able to
- Classify Claude's models and interfaces and choose the right one for each task
- Set up Projects with instructions and knowledge files so work does not start from zero
- Use Artifacts to create reports, tables, overviews and small interactive tools
- Write, test and roll out a custom Skill for a recurring task
- Set up connectors to your mailbox, calendar and file storage with deliberately limited access, and hand tasks to Cowork and routines
- Define team rules for data protection, review steps and shared templates
Course outline
Times show the standard running order and can be adjusted when we tailor the course to your team. The trainer stays available during breaks to help anyone finish an exercise.
Arrival and target picture 30 min
Where each participant stands today and which licences are available. Each person picks one use case to work on for both days. A realistic view of what Claude can do now and where its limits are.
Try it: Write your use case on a card: goal, input material and the result you want.
How language models work 60 min
Tokens and the context window, meaning how much text a model considers at once. Why the same question can give two different answers, and why models invent details and how to contain this. A matrix of Claude's model families by quality, speed and usage limit, and when Microsoft Copilot or ChatGPT is the better choice for a task.
Try it: Solve the same task with two different models and compare the results for quality and speed.
Good instructions and context 60 min
The building blocks of a good prompt: role, task, context, examples, format and limits, held together by the GCES method (Goal, Context, Expectations, Source). Letting Claude ask clarifying questions first, summarising and comparing long documents with supporting passages, and building review steps into the instruction.
Try it: Improve a weak prompt from your own work in three rounds and compare the versions side by side.
Improving two recurring templates 40 min
Taking two templates your team uses all the time, such as a standard reply and a recurring report, and rebuilding them with clear structure, fixed sections and built-in checks so they work every time.
Try it: Choose two templates, rewrite each with the three-round method, and keep the best versions for your Project.
Projects as shared memory 75 min
Splitting project instructions and knowledge files sensibly, and storing house style, product knowledge and templates centrally. Separating team and private areas so that sensitive content stays apart. Checking and correcting what Claude remembers, and switching memory off when it should not be used.
Try it: Create a Project for your use case with a set of instructions and three knowledge files, including your improved templates.
Artifacts, files and analyses 70 min
Reports, tables, charts and presentation drafts as Artifacts. Querying and comparing Excel, PDF and Word files, letting the analysis tool calculate figures and spot-checking them, and exporting, publishing and sharing Artifacts within the team.
Try it: Create an analysis with a chart and a short comment from an anonymised Excel list, then check two figures by hand.
Close 10 min
Recap of Day 1 and what to bring tomorrow.
Welcome back 10 min
Questions from Day 1 and the plan for today.
The ecosystem at a glance 55 min
Where things run: chat, the desktop app, Claude in Chrome, Cowork and Claude Code. What a Skill, a connector, MCP and a routine (a scheduled task) are. Agents explained simply as a workflow that combines AI with tools. A live demonstration from idea to finished Skill.
Try it: Assign your use case to the matching layer and justify the choice.
Writing your first custom Skill 45 min
A Skill's structure: a description of when it applies, the workflow, examples and templates. Writing and improving the Skill together with Claude, then adding it to your account.
Try it: Build a Skill for your use case, for example replies in your house style or a review report.
Testing, improving and sharing Skills 40 min
Testing a Skill on three real cases and tightening it. Sharing and maintaining Skills in a team: individual upload of a Skill file, and organisation-wide provisioning by an owner on Team and Enterprise plans, with named people responsible for each Skill.
Try it: Run your Skill on three real cases, improve it, and write a short note on who maintains it.
Connectors and connected services 60 min
Connecting Microsoft 365 or Google Workspace, file storage and calendar. Keeping permissions deliberately small, with read-only access first and approval before any action. Recognising hidden instructions in emails and documents (prompt injection).
Try it: Connect a connector and solve a task across mailbox and calendar, for example preparing for your next appointment.
Claude Cowork and routines 75 min
Handing over tasks instead of chatting, with access to a folder you release. Reading the plan, following progress and reviewing interim results. Scheduled tasks such as a morning overview or a weekly report, and the approval points before anything is sent or deleted.
Try it: Set up a routine for a daily overview and evaluate its first run together.
Team rules and your next 30 days 50 min
What may be uploaded and what may not. Maintaining shared Skills and Projects and assigning responsibilities. A reference sheet comparing the model families, kept up to date by your trainer, for you to take away.
Try it: Write a personal 30-day plan with three concrete steps.
Close 10 min
Recap of the two days and your plan.
What you'll need
Claude
- A Claude account on a paid plan (Pro, Team or Enterprise). Sharing Projects needs Team or Enterprise.
- Skills and Artifacts need code execution and file creation to be switched on; on Team and Enterprise this is an administrator setting. Provisioning a Skill organisation-wide is done by an organisation owner.
- Connectors for your mailbox, calendar and file storage, enabled by your administrator where your organisation allows them. If they are unavailable, the connector exercise runs as a demonstration.
- The Claude desktop app, recommended for the Cowork session.
Equipment
- A laptop with a current browser.
- A real use case from your daily work, and anonymised sample material for it.
Practice files are provided for every exercise. You are welcome to use your own everyday documents where your organisation's policy allows.