The starting point: everyday friction
The team already used AI tools. It wanted to put them to better use: preparing schedules from quotes, structuring campaign reports and avoiding the need to rebuild the same documents for each engagement.
The frustrations were specific. Generic responses. Sources that needed checking. An assistant that agreed instead of questioning an idea. Every experiment came with the same requirement: keep control over the quality of work delivered to clients.
Before the session, the agency collected its questions in a questionnaire. Strategy, project management, content, video and design were all represented. That document became the starting material for the workshop.
Turning questions into useful exercises
Nord Paradigm prepared Claude 101 around two areas: configuring the assistant for each person’s work and building reusable prompts for recurring tasks.
The session plan included a comparison before and after adding custom instructions, followed by building a prompt around a chosen task and examples. The method called for defining the deliverable, constraints and quality criteria before writing the instructions.
The prepared resources also addressed specific professional tasks: schedules, report analysis, SRT subtitles and After Effects expressions. Fictional files supported the planned demonstrations.
A quote defines the engagement. The work still needs a schedule.
The questionnaire asked for a way to produce a project schedule from a quote and kickoff information. The corresponding resource first asks for the start date, milestones, dependencies and periods to avoid.
It then requests a table connecting phases with tasks, deliverables, owners and required approvals. A weekly view of workload by role completes the request. The project owner remains responsible for checking whether the schedule is realistic.
The prompt’s value lies in that structure: it makes the required inputs and review points explicit before anyone uses the result.
A report needs to separate what is known from what is assumed
For campaign reporting, the team wanted to separate facts, hypotheses and recommendations. The prepared prompt follows that distinction directly.
Facts stay tied to the data. Possible explanations must be presented as hypotheses, with a description of how to check them. Recommendations must flag missing information. This structure supports human review; it does not remove the need for it.
Resources that remain available after the session
The workshop was accompanied by a dedicated page containing ten prompts to copy and adapt. It includes custom instructions, a method for building new prompts, report analysis and project schedules, alongside resources for files, interactive artifacts and video.
The prompt-building method asks the user to explain a task, provide appropriate examples and test the instructions on a fictional input. It gives the team a starting point for adapting the approach to other engagements.
The resources also address personal information, approved work environments and human review before delivery. For file operations, the prepared demonstration required a plan before making changes.
What this case demonstrates
The documented output is concrete: a workshop prepared around the team’s stated needs, demonstration materials and a prompt library tailored to agency work.
This case covers the workshop design and the resources made available. It does not measure time saved or adoption after the session. Those outcomes would require a separate follow-up.
For a team considering a workshop, it shows an approach that can be repeated: collect the tasks, define the expected deliverables and prepare resources people can use in their work.
The agency is not identified in this example. The situations described are drawn from the preparation questionnaire and workshop resources.