Episode 6 is here and this one puts AI to work in two different places.

There are two ways to turn your live report data into written narrative, automatically, and this episode covers both. 


        

            
        
    

The first is the Excel function, RW.Tools.AIInsight (available as of v5.0.0.91). Drop it in your template and it calls your AI provider at generation time, then writes the response into the workbook a named text box or a cell. 
That narrative becomes part of the Excel, PDF, or PowerPoint file you distribute. 
The syntax:

=RW.Tools.AIInsight( providerName, destination, prompt, [dataRange], [maxTokens] )

The first three are required the provider connection, where the answer goes (text box name or cell), and the prompt. 
On success the formula cell clears to empty and the narrative lands in your destination; if the provider can't be found or errors out, you'll see #VALUE!.

The second is AI Generated Insights on the Distribution step. Instead of writing into the file, it produces an AI Insight variable you embed in the email body, so the commentary travels with the delivery, around the report. 
You add a named prompt, pick your AI model, set the token limit and optional footer, write the prompt (you can drop variable tokens like {{Department}} right in), then insert the AI Insight variable into your message body. 
Every burst recipient gets commentary written from their own data. 

One prerequisite for both: an enabled AI provider connection (Settings ▸ AI Connections). Your admin sets this up once. 
Without it, the function call is skipped and the email section stays hidden. 

One file attached:
RW.Tools.AIInsight_Examples 

Unlike previous posts where we give you a file with working examples, this file is more a guide to show you configuration and format examples, use it to help you create your first AI driven insights.  

Here are some tips for writing effective commentary prompts:
1. Assign a clear persona and audience.  Tell the AI who it is and who it's writing for.  The persona shapes vocabulary, depth, and analytical framing.  The audience shapes the tone and assumed knowledge.
2. Give the data context explicitly.  The AI only knows what you tell it.  Explain what the table values it is reviewing is used for for ex: table shows monthly actuals vs budget by cost center and positive variances represent favorable under budget performance.
3. Be prescriptive about structure.  Unstructured prompts produce unpredictable formats ask for exactly what you want.  For example "respond with 3 bullet points one for revenue, gross margin, operating expenses following by a single closing sentence on overall outlook."
4. Tell it what to Prioritize.  "Focus on only the top 3 drivers of variance, ignore items under $50k variance."

Here are some things to avoid when writing commentary prompts:
1. Vague action words without a deliverable. Words like analyze, review, discuss give AI too much latitude.  Pair them with specific output.  
2. Asking why without supporting context.  Instead of "Explain why gross margin declined" a better way would be "Describe the gross margin result relative to budget and not the magnitude of the change".  This prevents AI from giving explanations without supporting context.
3. Avoid open-ended length instructions.  Instead of "Write a brief summary" try "Write a summary of no more than 100 words" this helps produce consistent results.

Here is a quick reference prompt checklist:
1. Does the prompt assign a persona and specify the audience?
2. Does the prompt describe what the data represents?
3. Is the output structure defined (bullets, paragraph, word count)?
4. Have you told AI what to prioritize or ignore?
5. Is your data range scoped to what the AI actually needs?
6. Have you set an explicit maximum output length?
Let us know what you think and if you have any questions!

• You need to login to have access to uploads.