I’ve seen several quality leaders complain this week about their disappointment with generative AI - they’re not getting the results they expected. And I understand why - context is king! If you just ask AI to write a procedure or generate a quality document, you’ll get generic, mediocre output. Without enough context, AI can only produce something generic based on its training data.

But how do you give it that context? By the time you’ve gone back and forth trying to “engineer the prompt” to get it to rewrite the same paragraph again, you’re left thinking “it would have been quicker to just write it from scratch!”

Was this email forwarded to you?

​

I’ve been experimenting with various generative AI tools a lot over the past year, and here’s one thing I’ve discovered that greatly improves writing with AI.

Don’t Just Use AI - Set Up True Collaboration

Think back to the last time you worked with a consultant or partnered with a colleague on a successful quality initiative. Did you just hand over a vague requirement and walk away? Probably not - you started with a conversation, a productive back-and-forth that built context and understanding. And if you had a good partner they probably asked you lots of questions.

The key is setting expectations from the start. Instead of just throwing requirements at AI, explicitly instruct it to ask you clarifying questions before generating content. This seemingly small change transforms how you interact with these tools.

Meta Example: How ‘We’ Created This Post

Let me show you what I mean using the creation of this very post as an example:

When I started drafting this article, I shared my initial thoughts with Claude (the AI chat I use). But instead of just saying “help me write this,” I specifically instructed it to:

“…ask me clarifying questions to make sure you understand my case and are meeting my goals before generating content for me.”

The result? Before suggesting any changes to my first draft, Claude asked:

• Who my specific audience was
• What action I wanted readers to take
• Whether I had specific examples to include
• If there was a particular reason I was writing about AI now

These questions forced me to clarify my own thinking and provided crucial context that generic instructions would have missed. Only then did Claude offer substantive suggestions that actually aligned with my goals. We went on to work on specific sections and the rather click-baity draft title.

It saved me time and effort, walking me step by step through the writing process rather than fighting against me. Crucially, this allows me to keep my voice in the writing, but also feels like less effort since I’m incrementally focussing on small pieces rather than struggling to rewrite a whole piece. There’s an added benefit that I don’t fill up the chat’s context window (and my own cognitive load) with useless rewrites.

Not only that, the first two questions Claude asked me are relevant to any marketing or writing content that I put together. So I added my answers to those questions to the "Project Knowledge" - in Claude that means any conversation I start under a particular Project will get the same starting context of prompt, documents and other information. A context that I continually improve as my interactions grow. But that’s a post for another day.

How to Set Up Your AI Collaboration

I believe this approach is so important that I specifically include versions of it in my project instructions:

“As an expert in X, you will ask clarifying questions to understand my requirements and objectives before generating content. When I provide drafts, you will discuss potential improvements, quality considerations, and possible issues before making changes.”

This transforms the AI from a basic tool into a thought partner - one that helps you explore options rather than just executing commands.

For those of us in quality, this should feel familiar - it mirrors the best practices we already apply in our work: clear requirements gathering, iterative review cycles, and continuous improvement. Just as you wouldn’t approve a procedure without proper review and questioning, don’t let AI generate content without the same rigor. This approach ensures your AI outputs maintain the same standards of quality you demand in every other aspect of your operations.

Try it out - you might be pleasantly surprised by how much more valuable your AI interactions become.

– Brendan

p.s. Enjoy this message? Read more at the Hyland Quality Systems website.