It is easy to write generally about AI tools. It is more useful to show how we actually use them. Here is a concrete look at how Claude from Anthropic is integrated into our working day, and what we have learned about when it works and when it does not.
What are Claude Projects?
One of the most important features in Claude for professional use is Projects: a system where you can create separate "projects" that have their own instructions and their own context memory across conversations.
In practice this means: we can set up a project for a specific client that contains background information about the business, the audience, tone of voice and previous content. When we then ask Claude for help with something for that client, we do not need to explain all the background again. Claude remembers it within the project.
For agency work this is a practical time-saver. For clients who write their own content, it is a way to use AI that actually knows the context.
What we use Claude for daily
Content drafts and editing We do not write finished client content with AI. But we use Claude to create a first draft we edit and rework, or to improve a draft we have written ourselves. It is faster and gives a better starting point.
Document analysis Claude can read a long RFP, a contract or a briefing and extract the most important points, identify what we should respond to and ask follow-up questions. For long documents this is a genuine time-saver.
Structured thinking When planning a new client engagement or thinking through a strategy, we use Claude as a thinking partner. We describe the situation and Claude helps us systematise thoughts, identify blind spots and see alternative approaches.
Translation and adaptation We publish content in multiple languages. Claude translates with a naturalness and understanding of context that is better than machine translation. We always edit, but the starting point is good.
Email and communication For complicated emails (where tone and precision really matter) Claude helps us find the right formulation faster.
What Claude does not replace
Most importantly: professional judgement. Claude knows a lot. It does not know what actually fits your specific client, your industry and your context. That is still human competence.
Original insight from experience. The best content we produce is based on things we have observed and learned through client work. AI cannot generate that.
Relationship-based communication. A follow-up conversation with a client, a difficult message that needs to land correctly: that requires human judgement and empathy.
Facts and current information. Claude has a knowledge cutoff and can be wrong. We always verify factual claims in content we publish.
Practical tip: get better results from Claude
The most important thing for good results: give good context. The more you explain who you are, who you are writing for and what you want to achieve, the better the result.
Example of a weak instruction: "Write a blog post about digital marketing."
Example of a strong instruction: "Write a blog post for small trade businesses about why they should ask existing clients to leave a Google review. The tone should be direct and practical, not academic. Maximum 600 words."
The second produces something usable. The first produces a generic article nobody will read.
Want to discuss how AI tools can be integrated into your working day in a way that actually saves time? Get in touch.