top of page

AI Powered Reporting Solutions: Copilot in Power BI, Licensing Needs and Prompt Design for Finance Teams

Writer: GrowthBI
GrowthBI
9 minutes ago
4 min read

Finance teams spend a large share of each month explaining numbers. Why did revenue fall in the North region? Why is freight cost over budget? What changed in working capital? Copilot in Power BI can help answer those questions faster, draft commentary and build visuals on request.


But Copilot is not magic. It needs the right licensing, a well-prepared data model and good prompts. This guide explains all three from a finance team's point of view.


1. What Copilot in Power BI can do


Copilot in Power BI uses large language models on top of your semantic model. Depending on your setup, it can:


  • Summarise a report page in plain English

  • Answer questions about the data in a semantic model

  • Create report pages and visuals from a description

  • Suggest DAX and explain existing measures for developers

  • Write narrative summaries that update with filters


For finance, the most useful features are question answering, summaries and narrative visuals for commentary.


2. Licensing needs


Copilot in Power BI is not included in a standard Pro licence on its own. Microsoft's current requirements are:


  • The report or model must sit in a workspace on a paid Fabric capacity (F2 or higher) or Power BI Premium capacity (P1 or higher)

  • A Fabric administrator must enable Copilot in tenant settings

  • Users need access to the underlying content, plus the appropriate Power BI licence to view or build

  • If your capacity is outside regions where Azure OpenAI is available, your admin may need to allow cross-region processing


Copilot consumes capacity units, so heavy use on a small capacity can slow other workloads. Monitor usage with the Fabric Capacity Metrics app.


Microsoft changes Copilot licensing and features often. Check current documentation before you plan your rollout.


3. Preparing your finance model for Copilot


Copilot reads your semantic model's names, descriptions and relationships. If those are unclear, answers will be too.


Use business names


Rename "Amt_LC" to "Amount (Local Currency)" and "GL_Acct_Desc" to "Account Name". Copilot understands plain language better than system codes.


Add descriptions


Add descriptions to key tables and measures. For example: "Net Revenue: revenue after returns, rebates and discounts, excluding GST."


Hide technical fields


Hide keys, sort columns and helper fields from report view so Copilot does not use them.


Define synonyms and linguistic schema


Add synonyms such as "sales" and "turnover" for revenue, or "FY" for financial year. This helps Copilot match the words people actually use.


Use a proper date table


Mark a date table and include your financial year and periods. Finance questions almost always involve time.


Prep data for AI


Power BI includes "Prep data for AI" features where you can set AI instructions, choose which fields Copilot should focus on and add verified answers to common questions. Use these to guide Copilot toward the right measures.


4. Prompt design for finance teams


The way you ask matters. Good prompts are specific about the metric, time period, filters and output.


Weak and strong prompts


Weak prompt

Strong prompt

How are sales going?

Show net revenue by month for FY2026 compared to budget, by division

Why are costs up?

Which expense accounts had the largest increase versus budget in Q1 FY2026?

Summarise this

Summarise this page in three bullet points for the CFO, focusing on variances over $50,000


Prompt patterns that work


  • Metric + period + comparison: "Gross margin % for last month versus the same month last year"

  • Metric + breakdown + threshold: "Operating expenses by cost centre where variance to budget is more than 10%"

  • Audience + format: "Write a short summary for the board in plain English, no more than 100 words"

  • Ranking: "Top 10 customers by revenue decline this quarter"


Prompts for commentary


"Summarise the main drivers of the revenue variance to budget for October by region and product family. Mention price and volume effects where visible. Keep it under 150 words."


Then review and edit. Copilot gives a first draft. The analyst owns the final words.


5. Validating AI output


Finance cannot afford wrong numbers in a board pack. Build checks into the process:


  • Reconcile key figures to the certified report before using Copilot's output

  • Test with known answers. Ask questions where you already know the result

  • Use verified answers for the most common questions

  • Keep a human in the loop for anything external or board-facing

  • Log errors and fix the model or descriptions that caused them


6. Practical uses for finance


  • Monthly management report summaries

  • Budget versus actual commentary drafts

  • Ad hoc questions from executives answered in minutes

  • Expense analysis by supplier, account or cost centre

  • Working capital trend explanations

  • Narrative visuals that update when filters change


7. How GrowthBI helps


GrowthBI helps finance teams in mid-market businesses with $30M to $500M in revenue get value from Copilot in Power BI. We prepare your finance model for AI, configure Copilot and train your team on prompts that give reliable answers.


Pricing


  • Project pricing: $5,000 to $30,000 AUD per project, depending on scope.

  • Ongoing support and improvements: from $2,000 per month.


Timeline


  • Most Copilot readiness projects take 1 to 6 weeks, depending on how much model preparation is needed.



Frequently Asked Questions


What licence do I need for Copilot in Power BI?


Content must be in a workspace on a paid Fabric capacity of F2 or higher or a Premium capacity of P1 or higher, with Copilot enabled by your admin. Users also need access to the content and the right Power BI licence.


Can Copilot write financial commentary?


Yes, it can draft commentary from your semantic model. A finance analyst should review, edit and approve it, especially for board or external reporting.


How do I improve Copilot's answers?


Use clear business names, add descriptions and synonyms, hide technical fields, set up a proper date table and use Prep data for AI features to guide Copilot.


Is Copilot accurate enough for finance?


It can be accurate when the model is well prepared and prompts are specific. Always reconcile key figures to certified reports and keep a human in the loop.


Does Copilot use our data to train Microsoft's models?


Microsoft states that Copilot in Fabric and Power BI does not use customer data to train its foundation models. Review Microsoft's current privacy documentation for full details.


Spend Less Time Explaining, More Time Advising


Copilot can take the first pass at finance questions and commentary. With a well-prepared model and good prompts, your team gets faster answers without losing control.


Book a free consultation with GrowthBI to get your finance reporting ready for Copilot.

 
 

Recent Posts

See All
bottom of page