Preparing for the workshop
Questionnaire for Course Preparation
Sebastian Goebel · Patent Attorney · Bösherz Goebel
Dear participants,
I am preparing the workshop you have signed up for. It is built around practice: by the end of the course everyone takes away something they can use in their own work. For that I need three things from you: which topics you want in the foreground, which software you already use, and what prior knowledge you bring.
The questionnaire has eight questions and takes about two minutes. There is no name field. I look at the answers of the whole group together and set the priorities from them.
Below the questionnaire you will find the preparation notes. Please check which section applies to you and work through it before the course day.
- Eight questions
- About two minutes
- Anonymous
For matching your answers to the session.
8 questions still open.
Setup Instructions
The three sections are ordered by level and build on each other. Work through the section that applies to you and skim the ones before it; we will look at the rest during the course.
If you attend online: connect a second monitor. On a single screen your own AI application covers the workshop content.
If you attend on site: bring your own laptop. If you want to try the local language models yourself, you need at least 16 GB of memory; with less, none of the models we use for that will run.
Getting started — chat in the browser
Create an account with one of the major providers and sign in once before the course day. A paid plan is useful for the course day: the free plans restrict access to the strong models we work with.
- Claude — Anthropic
- ChatGPT — OpenAI
- Microsoft Copilot — often already included in Microsoft 365
Further reading: The 4 levels of AI practitioner
Advanced — AI inside the workflow
At this level the tools work outside the chat window: on your files and inside your own workflows.
- Claude Code — command-line agent, carries out multi-step tasks
- Claude desktop app — the same agentic capabilities with a user interface and access to local folders
- Langdock — several models in one interface, hosted in the EU; for firms that provide access centrally
These tools read and change files on your computer. Please install the software for the exercises only on a computer that holds no confidential or important data, or isolate the installation in a sandbox.
Further reading: Assistance or integration · Multi-model workflows · Office + AI integration
Local AI — on your own hardware
The model runs on your machine and your input never leaves it. Install one of the programs below and download a model before the course starts. Depending on the model, the download takes minutes to hours.
- Ollama — lightweight, automation-friendly
- LM Studio — interface similar to ChatGPT
- Msty — RAG support, free for personal use
For the course day the laptop you bring is enough. For day-to-day use in the firm a dedicated machine for the local models is worth it, because they take up a lot of memory.
Further reading: Workshop: Local AI · Adopting local AI in a patent firm · Prompt engineering for local LLMs
Google Colab (optional)
For the advanced exercises I show examples in Google Colab. If you want to code along, you need a Google account and should open Colab once beforehand. It is not a requirement for the course: you can also just follow the examples on screen.
colab.research.google.comPlease study the e-learning part of the course beforehand.
Frequently asked questions
- Do I have to install local AI for the exercises?
- No. Cloud-based language models work just as well for the exercises. If you want to try local language models, bring a laptop with at least 16 GB of memory to an on-site course.
- Can I use my firm’s own language models?
- Yes. Every modern language model works for the exercises.
- What should I watch out for when installing agent-based systems such as Claude Code?
- These tools read and change files on your computer. Please install the software for the exercises only on a computer that holds no confidential or important data, or isolate the installation in a sandbox.
I look forward to the course day,
Sebastian Goebel, Founding Partner, Bösherz Goebel