Automate AI Reporting: Power BI & Gemini

Automate AI Reporting: Power BI & Gemini for Operations is a powerful tool designed to streamline workflows and boost productivity.

As Operations Managers, you know the daily grind of transforming raw operational data into actionable insights. Traditional reporting can be a time sink, often reactive, and rarely provides the predictive edge needed in today's fast-paced environment. This tutorial will guide you through integrating cutting-edge AI capabilities, specifically Google’s Gemini, with Microsoft Power BI to automate and enhance your operational reporting. We'll move beyond static dashboards to dynamic, intelligent operational intelligence that can pinpoint anomalies, forecast trends, and even draft narrative summaries, giving you back valuable time and providing deeper insights. This is the ai skill shift in practice, empowering you to make smarter, faster decisions.

Key Takeaways (TL;DR)

• Automate report generation and narrative summaries using Gemini AI for Power BI data.
• Leverage AI to identify key operational insights, trends, and anomalies from complex datasets.
• Create dynamic, interactive Power BI dashboards that are augmented by AI-driven text analysis.
• Reduce manual reporting effort and accelerate decision-making cycles within operations.
• Understand the practical integration steps for connecting Power BI with a large language model like Gemini.

Who This Is For & Prerequisites

This tutorial is designed for Intermediate Operations Managers and BI Analysts who are familiar with Power BI desktop, basic DAX, and the concept of APIs. We'll skip elementary definitions and focus on practical application.

Skill level: Intermediate Required tools/accounts:

• Microsoft Power BI Desktop: Latest version.
• Power BI Pro/Premium/Fabric License: For publishing and sharing.
• A Google Cloud Project with Gemini API Access: You'll need an active Google Cloud account and have the Gemini API enabled. Consider using Vertex AI for robust, enterprise-grade Gemini integration. Source: Google Cloud AI Documentation
• Basic understanding of Python: For scripting API calls (we'll provide snippets).
• Access to operational data: E.g., manufacturing KPIs, supply chain metrics, service desk performance, logistics data. Estimated time: 2-3 hours

What You'll Build/Achieve

You will build an intelligent operational reporting system that fetches data from a source (e.g., SQL Database, Excel, Dataverse), visualizes it in Power BI, and then uses the Gemini API to analyze key metrics. Gemini will provide natural language summaries, highlight critical changes, identify potential root causes, and even suggest next actions, all within your Power BI environment. This moves your reporting from mere data representation to automated insights and proactive operational intelligence.

---

---

Setting Up Your AI Environment: Google Cloud & Gemini API

Before we dive into Power BI, we need to ensure your AI assistant, Gemini, is ready to receive and process your operational data. This involves setting up a project in Google Cloud and enabling the necessary APIs. For enterprise deployments, leveraging Vertex AI for Gemini is the recommended approach, offering better control over model deployment, security, and scalability.

Step 1: Create a Google Cloud Project and Enable Vertex AI

First, you need a Google Cloud Project to manage your resources and APIs. If you already have one, you can use it.

1. Navigate to Google Cloud Console: Go to console.cloud.google.com. Log in with your Google account.
2. Create a New Project: From the project dropdown at the top, click "New Project."
   • Give your project a meaningful name (e.g., "Operations-BI-AI").
   • Select your billing account. If you don't have one, you'll be prompted to set one up (new accounts usually get free credits).
   • Click "Create."
3. Enable Vertex AI API: Once your project is created and selected, navigate to the Navigation menu (three horizontal lines) > Artificial Intelligence > Vertex AI.
   • You might see a prompt to enable the API. Click "Enable." This gives your project access to Gemini models via Vertex AI.

Step 2: Generate an API Key for Gemini Access

While Vertex AI typically uses service accounts for authentication, for a quick tutorial demonstrating API calls, an API key or setting up local credentials might suffice. For production, always use service accounts with granular permissions.

1. Navigate to IAM & Admin: In your Google Cloud project, go to Navigation menu > IAM & Admin > Service Accounts.
2. Create a Service Account:
   • Click "+ Create Service Account."
   • Give it a name (e.g., gemini-powerbi-svc).
   • Grant it the role of Vertex AI User (or a more specific role if you understand the permissions). This allows it to interact with Vertex AI.
   • Click "Done."
3. Create a Key for the Service Account:
   • Find your newly created service account in the list.
   • Under "Actions" (three vertical dots), select "Manage keys."
   • Click "ADD KEY" > "Create new key."
   • Choose "JSON" as the key type and click "Create."
   • A JSON file containing your service account credentials will download. Keep this file secure! This file will be used to authenticate your Python script.

Pro Tip: Never embed API keys or service account credentials directly into your Power BI files or public code. Use environment variables or secure credential management systems. For Power BI, the Python script will load this JSON file locally or use environment variables.

---

Preparing Your Operational Data in Power BI

The quality of your AI-generated insights directly depends on the quality and structure of your input data. Power BI is excellent for connecting, transforming, and modeling this data. Ensure your operational data is clean, well-structured, and contains the key metrics you want Gemini to analyze.

Step 1: Connect to Your Operational Data Sources

For this tutorial, let's assume you have a dataset detailing daily operational performance, such as:

• Date
• ProductionLineID
• UnitsProduced
• DefectRate
• DowntimeHours
• MachineMaintenanceCost
• LaborHours
• OrdersProcessed
• OnTimeDeliveryRate

1. Open Power BI Desktop: Start a new report.
2. Get Data:
   • Click "Get Data" from the Home tab.
   • Choose your data source (e.g., "SQL Server database," "Excel Workbook," "OData feed," or "Dataverse"). For simplicity, an Excel workbook with daily operational metrics is a great starting point.
   • Select your source and load the necessary tables.
3. Transform Data (Power Query Editor):
   • Click "Transform Data" after loading your tables.
   • Rename Columns: Ensure columns are clearly named (e.g., UnitsProduced instead of P_Units).
   • Data Types: Verify all columns have the correct data types (e.g., Date column as Date, UnitsProduced as Whole Number, DefectRate as Decimal Number).
   • Handle Missing Values: Decide on a strategy for nulls (e.g., replace with zero, average, or remove rows if appropriate).
   • Clean and Standardize: Ensure consistency in text fields (e.g., ProductionLineA vs production line a).
   • Create Calculated Columns (if needed): For example, Efficiency = UnitsProduced / LaborHours.

Step 2: Build Your Data Model

A well-structured data model will make your Power BI reports efficient and your AI prompts more effective.

1. Establish Relationships: In the Model view (the third icon on the left navigation bar), connect your tables using appropriate keys. For a single operational metrics table, this might be simple, but for multiple tables (e.g., OperationsData linked to MachineInfo and EmployeeData), define relationships (e.g., ProductionLineID to ProductionLineID).
2. Create Measures: Use DAX to create key performance indicator (KPI) measures. These are the values you'll be feeding to Gemini.
   • Example 1: Total Units Produced (Measure)

     Total Units Produced = SUM('Operational Metrics'[UnitsProduced])
     

   • Example 2: Average Defect Rate (Measure)

     Avg Defect Rate = AVERAGEX('Operational Metrics', 'Operational Metrics'[DefectRate] / 'Operational Metrics'[UnitsProduced])
     

   • Example 3: Monthly Downtime (Measure)

     Monthly Downtime = SUM('Operational Metrics'[DowntimeHours])
     

   These measures will be the numerical core of our AI analysis.

Insight: By creating measures beforehand, you ensure consistent calculations across your report and provide Gemini with aggregated, business-ready figures rather than raw, granular data. This is crucial for obtaining meaningful operational insights.

---

Connecting Power BI to Gemini via Python Scripting

Power BI can execute Python scripts directly, allowing us to leverage the Gemini API. We'll use Python to send our prepared operational data and a specific prompt to Gemini, then receive and parse the AI's response.

Step 1: Install Python and Required Libraries

If you don't have Python installed, download it from python.org. Ensure you check "Add Python to PATH" during installation.

1. Install Google Cloud AI Library: Open your terminal or command prompt and run:

   pip install google-cloud-aiplatform google-generativeai
   

   google-cloud-aiplatform is for Vertex AI interactions, and google-generativeai is for the public Gemini API. We'll primarily use google-cloud-aiplatform for enterprise use cases.
2. Install pandas: Power BI's Python integration uses pandas DataFrames.

   pip install pandas
   

3. Place Service Account Key: Make sure the JSON key file you downloaded in Section 1, Step 2 is in a secure, accessible location on your machine. You'll reference its path in the script.

Step 2: Create a Python Script to Call Gemini

This script will take a DataFrame (from Power BI) and a prompt, send it to Gemini, and return the AI's analysis.

1. Open Power BI Desktop Settings: Go to File > Options and settings > Options.
2. Navigate to Python scripting: Under the "GLOBAL" section, select "Python scripting."
3. Configure Python Home Directory: Ensure Power BI is pointing to your Python installation. If not, browse to your Python executable.

Now, let's write the Python script block. Create a blank query in Power BI (Get Data > Blank Query) or add a new source Script to get all your data from Python if you are directly querying it from Python. But for this tutorial, we will be passing data from Power BI to a Python script that will communicate with Gemini.

Here's a sample Python script that would reside within Power BI's "Run Python script" transformation:

import pandas as pd
import os
import vertexai
from vertexai.preview.generative_models import GenerativeModel, Part
from google.oauth2 import service_account


SERVICE_ACCOUNT_KEY_PATH = "C:/Users/YourUser/path/to/your-service-account-key.json"
PROJECT_ID = "your-google-cloud-project-id" # e.g., "operations-bi-ai"
LOCATION = "us-central1" # Or your preferred Vertex AI region


try:
    credentials = service_account.Credentials.from_service_account_file(SERVICE_ACCOUNT_KEY_PATH)
    vertexai.init(project=PROJECT_ID, location=LOCATION, credentials=credentials)
    model = GenerativeModel("gemini-pro") # Or "gemini-1.0-pro"
except Exception as e:
    print(f"Error initializing Vertex AI or loading credentials: {e}")
    # Return an error DataFrame if initialization fails
    output_df = pd.DataFrame([{"AI_Analysis": f"Error: Could not initialize AI. {e}"}])
    exit() # Exit the script early


data_string = dataset.to_csv(index=False)


prompt_text = f"""
As an Operations Manager, analyze the following daily operational performance data.
Focus on identifying trends, anomalies, potential root causes, and providing actionable recommendations to improve efficiency and reduce costs.
Highlight any significant deviations from the previous day or average performance.
Data (CSV format):
{data_string}

Provide a concise summary, key insights, and specific operational action items.
Format your output as a narrative, suitable for a daily operational report.
Include a specific header for "Key Insights" and "Action Items".
"""

try:
    # Send the prompt to Gemini via Vertex AI
    instances = [Part.from_text(prompt_text)]
    response = model.generate_content(instances)

    # Extract the generated text
    ai_analysis = response.candidates[0].content.parts[0].text

    # Create a pandas DataFrame from the AI's response
    # This DataFrame will be passed back to Power BI
    output_df = pd.DataFrame([{"AI_Analysis": ai_analysis}])

except Exception as e:
    print(f"Error calling Gemini API: {e}")
    # Return an error DataFrame if API call fails
    output_df = pd.DataFrame([{"AI_Analysis": f"Error: Failed to get AI analysis. {e}"}])

Step 3: Embed the Python Script in Power BI

Now, let's integrate this Python script into your Power BI report.

1. Load Your Operational Data: Ensure you've loaded your operational data and created the necessary measures, as detailed in Section 2.
2. Select Data for AI Analysis: In Power BI Desktop, go to the "Data" view (table icon on the left).
   • You'll need a table or a query that contains the aggregated data you want Gemini to analyze.
   • For instance, if you want daily summaries, ensure your data is grouped by Date and includes your key measures.
   • Select the table/query containing the data you want to send.
3. Run Python Script Transformation:
   • In the Query Editor (Transform Data > Transform Data), select the query/table you want to pass to Python.
   • Go to the "Transform" tab and click "Run Python script."
   • Paste the Python script from Step 2 into the script editor window.
   • Crucially, ensure the input DataFrame is named dataset (as in the Python script). Power BI automatically passes the selected table as a DataFrame named dataset to the Python script.
   • Click "OK."
   • Power BI will execute the script. If successful, a new table will appear in your Query Editor, containing the AI_Analysis column returned by the output_df.
   • Rename this query: Rename it something clear, like Gemini_Operational_Insights.
   • Click "Close & Apply" to load this new table into your Power BI data model.

Critical Note: For passing different data subsets to Gemini (e.g., daily metrics vs. weekly trends), you'll need to create separate queries or parameters in Power BI that feed specific dataframes to your Python scripts.

---

Crafting Effective Prompts for Operational Insights

The quality of Gemini's output is highly dependent on the quality of your prompt. As Operations Managers, you need insights that are actionable and relevant to your domain. This requires careful consideration of what questions you want the AI to answer.

Step 1: Understand Prompt Engineering Fundamentals

Think of the prompt as giving instructions to a new, highly intelligent intern who understands operations but needs specific guidance on what to look for and how to present it.

• Be Clear and Concise: Avoid ambiguity.
• Provide Context: Explain the role (Operations Manager) and the type of data.
• Define Output Format: Specify headings, bullet points, narrative style.
• Specify Desired Insights: What exactly do you want to know? (Trends, anomalies, root causes, recommendations).
• Iterate and Refine: Your first prompt won't be perfect. Test and improve.

Step 2: Examples of Operational Prompt Structures

Let's refine the prompt from the Python script with specific operational targets.

Example 1: Daily Performance Review

"Analyze the attached daily operational performance data for a manufacturing line.
The data includes 'UnitsProduced', 'DefectRate', 'DowntimeHours', and 'OnTimeDeliveryRate'.
Your audience is an Operations Manager.

**Context:** This data represents yesterday's shift performance. Compare today's metrics with the previous day and the 7-day average.

**Specific Tasks:**
1.  **Summarize Overall Performance:** Provide a high-level overview.
2.  **Identify Key Deviations:** Pinpoint any metric that deviated more than 5% from the previous day or 10% from the 7-day average. For each deviation, suggest 1-2 potential operational root causes (e.g., machine failure, material shortage, staffing issue).
3.  **Propose Action Items:** Based on the deviations and potential causes, suggest 2-3 specific, actionable steps an Operations Manager should take *today*.
4.  **Forecast (Briefly):** Based on the current trend, briefly mention if performance is likely to improve or decline in the next 24 hours.

**Output Format:**

## Executive Summary
[2-3 sentences]

## Key Insights & Deviations
-   [Metric Name]: [Current Value] (vs Previous: [X%], vs 7-day Avg: [Y%])
    *   **Potential Causes:** [Why this happened]
-   ...

## Recommended Action Items
1.  [Action Item 1]
2.  [Action Item 2]
3.  ...

## Short-term Outlook
[1-2 sentences]

Data (CSV format):
{data_string}
"

Example 2: Inventory Optimization Suggestion (Requires including inventory data in data_string)

"Analyze the attached inventory data, focusing on stock levels, daily consumption rates, and lead times for critical components (Part A, Part B, Part C).
Your audience is a Supply Chain & Operations Manager.

**Specific Tasks:**
1.  **Identify Overstock/Understock Risks:** Highlight any components (Part A, B, C) where current stock is projected to run out within 5 days OR is more than 30% above the optimal safety stock level.
2.  **Suggest Reorder Points:** For understocked critical components, recommend an urgent reorder quantity based on current consumption and lead times.
3.  **Identify Slow-Moving Inventory:** List any components with no consumption in the last 30 days and suggest potential actions (e.g., promotional sale, return to vendor, re-evaluate forecast).

**Output Format:** Narrative with clear headings for each section.

Data (CSV format):
{data_string}
"

Best Practice: Start with simpler prompts and gradually add complexity. Test different data subsets to see how Gemini responds. The more specific your constraints and desired output, the better the result.

Step 3: Handling Data Volume in Prompts

Gemini has token limits. For very large datasets, you might need to:

• Pre-aggregate in Power BI: Summarize data to a higher level (e.g., daily aggregates instead of hourly raw sensor data).
• Focus on Key Metrics: Only send the most critical numbers for analysis.
• Chunk the Data: If analyzing long duration trends, send data in chunks or focus on summary statistics over time.

---

Integrating AI-Generated Narratives into Power BI Dashboards

Now that you have your AI-generated insights returning to Power BI as a table, the next step is to display them effectively on your dashboards. This is where ai operations reporting truly shines.

Step 1: Create a Dedicated AI Insights Visual

You'll typically want a text-based visual to display Gemini's narrative output.

1. Add a Card or Text Box Visual:
   • In Power BI Desktop, on your report page, click the "Card" visual icon or "Text box" icon from the "Visualizations" pane.
2. Display the AI Analysis:
   • If using a Card visual: Drag the AI_Analysis column from your Gemini_Operational_Insights table into the "Fields" well. Power BI will automatically display the first (and likely only) value.
   • If using a Text box visual: You can manually insert a field reference using "Add dynamic content" or by typing your text and inserting the AI_Analysis field. This offers more formatting flexibility.
3. Adjust Formatting:
   • Resize the visual to accommodate the text.
   • Format font size, color, background, and borders to match your dashboard's theme. Ensure readability.
   • You might want to add a clear title like "Gemini AI Operational Insights" above the visual.

Power BI Feature Tip: For more control over AI narrative formatting, consider creating a custom visual if you have development skills, or using a "Smart Narrative" visual and augmenting it manually, though our goal here is full AI automation.

Step 2: Use Slicers to Drive Dynamic AI Analysis

The real power comes when your AI insights respond dynamically to user selections within Power BI.

1. Add Date Slicers: Add a "Slicer" visual to your report.
   • Drag your Date column (from your primary operational data table) into the slicer's "Field" well.
   • Configure the slicer (e.g., "Relative date" for "Last 7 days" or "Between" for a date range).
2. Connect Slicer to Query (if not automatic):
   • Ensure your date slicer filters the main operational data table.
   • When the main operational data table is filtered by the slicer, the data passed to your Python script (via the "Run Python script" transformation) will also be filtered. This means Gemini will analyze only the visible, filtered data.
   • This setup allows Operations Managers to select a desired time period and instantly get an AI-generated summary for that specific period.

> Critical Integration: The key is that the Power BI query feeding into your Python script must react to the report's filters. If your "Run Python script" step is based on an aggregated table that *doesn't* get filtered by your date slicer, your Gemini analysis won't be dynamic. Ensure your data flow allows filter propagation.

Step 3: Combine AI Insights with Traditional Visuals

Integrate the AI narrative with your existing KPIs and charts.

• Contextual Placement: Place the Gemini_Operational_Insights visual alongside key performance charts (e.g., Units Produced by Line, Defect Rate Trend). This allows users to see the raw data and then immediately read the AI's interpretation.
• Conditional Formatting: Consider using Power BI's conditional formatting on your core metrics. If Gemini highlights a specific issue (e.g., "Defect Rate increased by 10%"), visually emphasize that metric on your dashboard (e.g., red color for high defect rate).

---

Automating Refresh and Enhancing Operational Intelligence

The goal of ai operations reporting is automation. You want your AI insights to be generated and updated automatically without manual intervention. This takes your reporting from reactive to proactive, providing true operational intelligence.

Step 1: Schedule Dataset Refresh in Power BI Service

For your reports to be constantly up-to-date, the underlying datasets in Power BI Service must refresh regularly.

1. Publish Your Report: Save your Power BI Desktop file (.pbix) and publish it to a Power BI Workspace in the Power BI Service.
2. Configure Gateway (if applicable): If your data sources are on-premises (e.g., on-premise SQL Server, Excel files on a network drive), you'll need to configure a Power BI On-premises Data Gateway. This gateway allows Power BI Service to securely connect to your local data.
3. Schedule Refresh:
   • In Power BI Service, navigate to your Workspace.
   • Find the dataset corresponding to your published report.
   • Click the three dots next to the dataset > "Settings."
   • Expand "Gateway connection" and ensure it's configured correctly.
   • Expand "Scheduled refresh." Turn on scheduled refresh.
   • Add refresh times (e.g., daily at 6 AM, or multiple times a day). This will trigger the Power BI data refresh, which in turn re-executes your Python script for the latest data.
   • Crucial: Ensure your Python script (and thus Gemini API call) can complete within the refresh time limit and that you have sufficient API quota.

Step 2: Monitor API Usage and Cost

Using external APIs incurs costs. For Vertex AI Power BI integrations, monitor your Google Cloud project.

1. Google Cloud Monitoring: In your Google Cloud Console, navigate to Vertex AI > Monitoring or APIs & Services > Dashboard and Quotas.
2. Set Up Alerts: Configure budget alerts in Google Cloud to notify you if your Gemini API usage exceeds a certain threshold. This helps prevent unexpected bills.

Step 3: Implement Version Control and Review

As your prompts evolve and your operational needs change, manage your Python scripts and Power BI reports effectively.

• Version Control for Scripts: Store your Python scripts in a version control system (e.g., Git) outside of Power BI Desktop. This allows tracking changes, collaboration, and easy rollback.
• Regular Review of AI Outputs: Periodically review Gemini's generated narratives. AI models can sometimes "hallucinate" or provide less accurate insights based on data nuances. Your operational expertise is vital for validating these outputs. This continuous feedback loop helps you refine your prompts over time.

---

Expected Results

Upon successful completion of this tutorial, you will have:

• A Power BI report featuring standard operational dashboards and KPIs.
• A dedicated section or visual on your dashboard displaying dynamic, AI-generated narrative insights from Google's Gemini.
• These insights will summarize operational performance, flag anomalies, suggest potential root causes, and propose actionable recommendations tailored to the selected data period.
• The entire process, from data fetching to AI analysis, will refresh automatically on a schedule, providing continuous automated insights into your operations.

How to verify it worked:

1. Open your published Power BI report in the Power BI Service.
2. Use a date slicer to change the time period (e.g., "Last 7 days" to "Last 30 days").
3. Observe the AI insights visual: the text should dynamically update to reflect the newly filtered data, offering a fresh analysis.
4. Check the content of the AI narrative for relevance, accuracy, and actionable advice specific to your operational metrics. Compare it to manual analysis you might perform.

---

Troubleshooting

Common Issue 1: "Python script error" in Power BI

Cause: This is broad and can be due to Python environment issues, missing libraries, or errors in your script. Solution:

1. Check Power BI Python Path: Go to File > Options and settings > Options > Python scripting. Ensure Power BI is pointing to the correct Python installation where you installed google-cloud-aiplatform and pandas.
2. Verify Library Installation: Open your command prompt/terminal and run pip list to confirm google-cloud-aiplatform and pandas are listed.
3. Debug Python Script Locally: Copy your Python script content into a standalone .py file. Replace dataset = pd.DataFrame(...) with a dummy DataFrame mimicking what Power BI passes. Run the script directly from your terminal (python your_script.py) to catch syntax errors or API communication issues.
4. Error Messages: Pay close attention to the error message Power BI provides; it's often very specific about the line number or type of error. Common errors include:
   • ModuleNotFoundError: Missing google_cloud_aiplatform or pandas.
   • Authentication Error: Problem with your service account key file path or credentials.
   • API Quota Exceeded: You've hit the Gemini API usage limit.

Common Issue 2: AI_Analysis column is empty or contains an error message

Cause: Gemini API call failed, or the response couldn't be parsed. Solution:

1. Check Internet Connectivity: Ensure the Power BI machine (or Gateway server for scheduled refreshes) has internet access to Google Cloud APIs.
2. Verify Google Cloud Project ID and Location: Double-check PROJECT_ID and LOCATION variables in your Python script against your Google Cloud project settings.
3. API Key/Service Account Credentials: Re-verify the SERVICE_ACCOUNT_KEY_PATH is correct and the JSON file is valid and accessible by the Python script (permissions). The service account needs Vertex AI User role.
4. API Quota: In Google Cloud Console, check APIs & Services > Quotas for the Vertex AI API. Ensure you haven't exceeded any limits. Request an increase if necessary.
5. Gemini Model Availability: Ensure the model name (gemini-pro or gemini-1.0-pro) is correct and available in your chosen region.
6. Prompt Length: If your data_string is very long, it might exceed Gemini's token limit. Reduce the data sent or refine your prompt for conciseness.

Common Issue 3: AI insights are not dynamic when slicing data

Cause: The Power BI query feeding the Python script isn't being filtered by your dashboard slicers. Solution:

1. Inspect Query Dependencies: In Power BI Query Editor, go to View > Query Dependencies. Trace the lineage of your Gemini_Operational_Insights query back to your raw data.
2. Apply Filters Correctly: Ensure that the query used as the input dataset for your "Run Python script" step is indeed impacted by the report-level filters or slicers. You might need to create an intermediate query that applies filters before passing the data to Python. For example, if you want daily filtered data, your Python script input should be the filtered daily data table, not a static summary.

---

Automate AI Reporting: Power BI & Gemini for Operations is ideal for teams that need faster execution and measurable outcomes.

Pricing context (USD): Teams typically spend $20-$100 per user/month depending on plan and usage.