Short answer: Choose a workshop built around a real business task, demand practice with a deliverable, and decide in advance who will keep using it after the workshop.
A goal like "get to know AI" is too broad to tell whether the workshop succeeded. A goal like "build a draft price quote based on our services" lets you check a result.
What do you prepare beforehand? One process, an example that contains no sensitive information, and a person from the business who knows the work. You can work with a dummy document or data stripped of identifying details. These are practice materials, not approval to upload private information to an external tool.
What should happen during the workshop? A short explanation, a demonstration, and then the participants do it themselves. The business owner should know how to fix the deliverable and understand what requires human review, not just watch a lecturer type.
Which deliverable counts as success? A template for a price quote, a draft answer to repeating questions, or a small process you can repeat. A demo website is a prototype; publishing a business website requires additional checks of content, forms, accessibility and security, depending on the use.
How do you vet the instructor? Ask for an example of an exercise and of the deliverable a participant builds. A lecture can fit exposure to ideas, and a workshop fits practice. Choose by the goal, not by the number of tools that appear in the slides.
In my Shafayim talk I split the time into 45 minutes of explanation and 45 minutes of live demo building websites from the audience's ideas. That is the format I described in the original post, not a rule every workshop must copy. My quote from there sums up the direction: "All that's left for you is to try."
What do you do the week after? Go back to the chosen task, use the deliverable, and write down what worked and what still needs fixing. If nobody opens the deliverable again, check whether the right task was chosen and whether instructions are missing.
I help businesses choose an existing tool, install it, and train the team around a useful process. Details: https://israeli-ai-agency.pages.dev/services/automation/ .