Prompting Arbor AI and Ask Arbor (the Basics)

This guide serves as a cheatsheet to help you chat with (prompt) Arbor's AI features, like Ask Arbor, in such a way that answers are more accurate and with more AI credit efficiency. For a more detailed exploration of prompting, please see our Prompt Like a Pro - Ask Arbor and Arbor AI article.

How to talk to Arbor AI 

Arbor's AI features are tools, not humans. While it feels like a conversation with Ask Arbor, for example, it's not a true two-way street. You can't just speak your mind and expect the AI to fill in the blanks. Vague prompts lead to generic answers and a lot of back-and-forth, which leads to wasted time and credits. So, learning to effectively speak to Ask Arbor is crucial.

Using Ask Arbor starter prompts

To make it easier to discover what Ask Arbor can do, we have built a number of out-of-the-box prompts you can use to instruct Ask Arbor to take specific types of actions. You can Ask Arbor to: Find, Today, Draft, Trends, Act, and Report.
 

Each of these categories represents the actual outcomes of a prompt you want Ask Arbor to achieve. Simply select what you want Ask Arbor to do, and Ask Arbor will go back and forth with you untill it has enough information to accurately honour your request.
 

Using these pre-built prompts helps reduce the time you spend figuring out what the system can do, and helps you immediately get accurate data without having to type everything from scratch.

Writing your own prompts

Whether you already have a good idea of what Ask Arbor can do, you are using an AI feature without starter prompts (like Suggest Email), or you simply want to submit a request on your own terms, you can prompt Arbor AI with your own words and prompts. 

Successful prompts are clear and specific; they do not leave room for ambiguity or for the AI to "fill in the blanks". To help with this, we advise avoiding tentative questions; be clear and commanding instead. Tell the AI what to do, rather than asking the AI if it can do something.

The basic formula

To help get you started with this and construct powerful, direct commands, use the VEST formula: [VERB] + [ENTITY] + [SPECIFICS] + [TIMEFRAME] 

• V (Verb): Specify the action you want the AI to take (e.g., Show me, Draft, Calculate).

• E (Entity): Define the subject of your request (e.g., Attendance, Behaviour Incidents).

• S (Specifics): Apply any filters or conditions (e.g., for Pupil Premium students).

• T (Timeframe): Set the time parameters (e.g., this week, last term).

Applying the Formula:

A well-constructed prompt combines these elements for clarity and precision. For example, instead of a vague question, follow this structure to build a strong command.

• Incorrect: "What's the status of attendance and stuff for Year 7 recently?" (This prompt lacks directness and specificity).

• Correct (Command Formula): "Show me [VERB] the attendance [ENTITY] for Pupil Premium students [SPECIFICS] this week [TIMEFRAME]" (This command gives the AI clear instructions on the action, subject, filtering, and timeframe needed).

Once you are comfortable with this basic formula, you can further enhance your prompts by adding details around goals, formatting needs, specific context, and tone, referencing the visual for additional inspiration. This structured approach helps ensure you get the exact results you need efficiently.

For a more detailed exploration of prompting, please see our Prompt Like a Pro - Ask Arbor and Arbor AI article.

Conversing with Ask Arbor

Ask Arbor remembers the context of previous messages within the same chat. This means you can chain your prompts and follow up on Ask Arbor's responses, instead of restarting from scratch. To get confident in prompting Ask Arbor, back-and-forth conversations are encouraged as these give you options to refine your requests, such as adding more details, or to validate if information is correct by asking follow-up questions like commanding Ask Arbor to explain why or how it came to a certain conclusion.
 

Equally, if your request to Ask Arbor is vague or incomplete, Ask Arbor will challenge you and request clarifying information to ensure you get an accurate and helpful outcome instead of half-guessing at an answer.

For example, if you ask "Would you be able to report on attendance for me?", Ask Arbor will hone in on what you mean exactly before outputting a response.
 

What to avoid

Vague and Tentative Prompts

The AI does not know your personal or professional context, unless you provide it. This is where the principle of "garbage in, garbage out" comes into play. If your prompt is vague, the AI's response will be vague. The AI is only as good as the prompt you give it. 

Furthermore, do not phrase your prompts as tentative questions (e.g., "Can you show me..."). Use the VEST formula instead, for example.

Not confident enough in writing your own well-structured prompts yet? No problem, use one of our many starter prompts and have Ask Arbor walk you through the process.

Niche Date Ranges

There are certain niche date ranges Arbor AI can't do right now. For example, asking for data for "every Friday". Instead, stick to the types of date ranges you might set for a custom report.

Subjective Adjectives

Avoid subjective adjectives like “wins/losses, best/worst, effective/ineffective, large/small, high/low, broad/narrow, significant/insignificant, improved/deteriorated, reliable, simple, intuitive, efficient, robust, etc." 

The AI does not have the context to know what "bad attendance" is, for example, so leaving the definition of bad open to interpretation is likely to lead to unreliable and inconsistent results.

Instead, make sure to define subjective things. In the case of "bad attendance", you get better results if you explain what your threshold for bad is. For example:

• You might say something like "show me students with statutory present attendance for the year under 85%" 
• Avoid asking "show me students with bad attendance" 

Leading Questions

Avoid leading questions that attempt to guide the AI's response, such as “Isn't it better to…”. Leading questions have a tendency to push an AI toward a specific point of view and can hamper objective responses and consistency of replies.