Write clear prompts with action, content scope, and intent

Why prompts fail

Here’s a prompt that seems reasonable: “Help with this article’s SEO.” The AI Agent receives this and needs to guess: Which article? What kind of SEO help - generate metadata, analyze keywords, update existing fields? Should it make changes directly or show you options first? Every time the agent has to guess, you risk getting results that don’t match what you need. And even when the agent eventually gets it right, vague prompts lead to extra back-and-forth that takes more time and uses more resources. The solution is building prompts from three essential principles that eliminate guesswork.

The three essentials

Every prompt you write to the AI Agent should answer three questions:

1. What action do you want the AI Agent to take?
2. Which specific content should it work with?
3. What’s your intent - do you want suggestions to review, or direct implementation?

Think of these as building blocks. Miss one and the AI Agent has to fill in the gaps on its own. Let’s now look at each principle closely so you can give the AI Agent the information it needs to deliver valuable results.

Principle 1: State the action clearly

Start your prompt with a specific verb that describes exactly what you want done. Vague requests like “help with this article” or “fix the SEO” force the AI Agent to interpret your intent, often inefficiently or incorrectly. The strongest action verbs describe concrete content operations. You can find a list of such action verbs in the table below. Let’s now go over a few examples to show how specific verbs transform your requests. Example 1: You want to adjust the meta description of an article you’re currently working on.

• ❌ Vague: “Help with the description for our campaign”
• ✅ Clear: “Generate a meta description for the Article item named Spring Campaign 2026”

Example 2: Another content piece needs to be translated into German.

• ❌ Vague: “I need the onboarding guide in German”
• ✅ Clear: “Translate the Getting Started Guide content item to a new German language variant”

Principle 2: Specify the content

After stating the action, tell the AI Agent exactly which content to work with using Kontent.ai terminology. Being specific about scope prevents the AI Agent from affecting too much content or too little.

How to specify scope

Let’s look at some example prompt fragments identifying the scope using the Kontent.ai terminology for precise results. Single item by name
“...the content item named Spring Campaign 2026” Single item, specific language
“...the German language variant of Product Overview” All items of a type
“...all Article content items” Items in a collection
“...all content items in the Blog collection” Items in a workflow step
“...content items in the Ready for Review workflow step” Combined filters
“...all Blog Post content items in the Marketing collection that are in the Draft workflow step”

Common terminology mix-ups to avoid

Principle 3: Declare your intent

The third essential tells the AI Agent whether you want suggestions for human review or direct action. This is your safety mechanism to ensure changes happen only when you’re ready.

1. Suggest first when you want control. Use phrases like “Suggest taxonomy terms for…” or “Recommend workflow changes for…” or “Show me options for…”. This gives you a chance to review before anything changes.
2. Request direct action when you’re confident. Use phrases like “Create a new content item…” or “Update the pricing element…” or “Assign these taxonomy terms…” This tells the AI Agent to execute immediately.

Think of it this way: the more items affected or the more important the change, the more valuable the “suggest first” pattern becomes. For single, low-risk changes, direct action works fine. For anything touching multiple items or affecting published content, start with suggestions.

Example conversation

Notice the language used in the prompts – action verb used, scope specified, and the use of Kontent.ai terminology. You: “Suggest taxonomy terms for the Article content item Cloud Migration Best Practices from the Topics taxonomy group.” AI Agent: Responds with recommendations. You: “Assign those terms as suggested.” This two-step pattern gives you control. You see what the AI Agent will do before it happens, and you can adjust if the suggestions don’t match your expectations.

Putting the three essentials together