Evaluating AI Visualisation tools in Power BI

Evaluating AI Visualisation tools in Power BI

The integration of artificial intelligence (AI) into data analytics tools like Microsoft's Power BI has ushered in a new era of efficiency and insight.

The existing AI features like Q&A, Decomposition Trees, and Smart Narratives have already made it easier to analyse data by letting users ask questions in plain language and get visualisations or written insights in return. This has also helped make data more accessible to non-technical users.

But the new Copilot AI assistant in Power BI is truly groundbreaking. Copilot uses advanced AI to automatically generate entire reports, provide data summaries, and assist with analysis - just by having a natural conversation with it.

In this blog, we'll dive into all the AI capabilities in Power BI, looking at what they can do, limitations to be aware of, and best practises for getting the most out of these powerful AI tools. By understanding AI's potential and how to use it effectively, businesses can gain valuable data insights more quickly and easily.

Existing AI Visual Tools

Q& A, Power BI

Q&A visual in Power BI is a type of conversational AI interface that allows users to ask questions about their data in plain English and get visual representations of the answers.

Pros:

• Power BI Q&A empowers users to explore data through natural language queries. It provides a variety of suggested questions to get you started, or you can type your own in plain English.
• It can understand many types of queries like filtering, aggregations, comparisons etc. Users can ask things like "show me total sales by region" or "what were the top 5 products last year?"
• Simply ask a question about your data, and Q&A will automatically generate visuals like cards, charts or tables to answer it. Users can turn a Q&A result into a standard visual.
• After Q&A displays your results, you can use the interactive features to keep the conversation going to uncover more insights.

Cons:

• Q&A might not understand all words you use and will highlight them in grey. It will still attempt an interpretation based on available data.
• It isn't always possible for Q&A to display the data using the visual type requested by the user. Instead, it can prompt you with a list of workable visual types.
• Be aware that the generated visuals are based on AI analysis and may require further validation for critical insights. Microsoft may use anonymised user queries to improve Q&A, following their privacy statement.

While the AI capabilities in Power BI can significantly enhance data analysis and insight generation, effectively communicating those insights remains crucial. Read our blog 3 key elements to get data storytelling right to learn how to perfect your data narratives.

Key Influencer

Key Influencers visual in Power BI allows you to analyse which dimensions or factors in your data are the biggest drivers or influencers of a specific metric or measure.

Pros:

Key Influencer visual is a powerful AI tool that helps you understand what truly impacts your chosen metric. It analyses your data, identifies the most important factors influencing that metric, and presents them in a clear ranking. By utilising Key Influencers, you can gain valuable insights into the root causes affecting your metrics and make data-driven decisions to optimise your results.

It can help:

• Identify Key Drivers: Discover the factors that have the greatest influence on your metric.
• Relative Importance: Compare these factors to each other in terms of their impact.
• Flexibility: Analyse both categorical (Point Biserial correlation tests) and continuous (Pearson correlation) factors depending on your target metric. When a linear relationship isn't clear, the tool might generate up to five data bins for better analysis.
• Detailed Analysis: The ‘Key influencers’ tab assesses each factor individually, while the ‘Top segments’ tab shows how a combination of factors affects the metric.

Cons:

• Data Connectivity: Key Influencers currently doesn't support analysis with data sources like Direct Query, Live connections to Azure Analysis Services and SSAS, or publishing to web or SharePoint.
• Limited Explanation: While the tool identifies influential factors, it might not always provide a detailed explanation for their impact. This can require further investigation on your part.
• Data Complexity Challenges: Key Influencers works best with data that has relatively clear and linear relationships. For highly complex datasets or intricate cause-and-effect chains, the insights may be difficult to interpret, or the analysis may not be as effective.

Decomposition Tree

The Decomposition Tree visual in Power BI is used to analyse a measure or metric by breaking it down across multiple dimensions or attribute hierarchies visually.

Pros:

When you want to analyse a measure or an aggregate, the Decomposition Tree's intuitive interface and AI capabilities help you navigate your data, find root causes, and gain valuable insights for decision-making.

• Drill Down Deeply: Breaks down your data into its components by adding dimensions layer by layer, allowing you to explore in any order. The AI automatically aggregates data and calculates splits, helping you identify the factors driving your chosen metric.
• High & Low Values: Switch on "Enable AI Splits" to see the factors contributing to high and low values within your data.
• On-Demand Summaries: Add the icon to the header of a visual; when hovered over, it provides report readers with instant summaries.

Cons:

• Data Size and Complexity: The Decomposition Tree is challenging to manage with large or highly complex datasets. It has limitations on the number of levels (50), top N (10) per level, and data points that can be visualised at one time on the tree (5000). Highly intricate datasets or complex measures might not be well-suited for this visual.
• Compatibility Considerations: Decomposition Trees are not supported for complex measures and are not currently supported for on-premises Analysis Services, Azure Analysis Services in some scenarios, within Q&A, or when publishing reports to the web.

As the AI capabilities in Power BI continue to evolve and transform data analysis, partnering with experts can ensure you maximise the potential of these powerful tools. Book a call with our data visualisation experts at Ei Square to guide your through the AI integration journey.

Smart Narratives

Narrative also known as Smart Narratives is a Power BI visual that automatically generates insightful narratives and descriptions about the data in natural language.

Pros:

• Smart Narratives provides concise text summaries of your visuals and reports. These summaries highlight key findings and offer insightful observations, helping your audience understand the data quickly and efficiently.
• It uses natural language generation (NLG) techniques to automatically analyse data patterns, trends, and anomalies, and generate insightful narratives in plain language. This eliminates the need for analysts to manually describe every data point, saving time and effort.
• It adapts and updates the narratives dynamically as users interact with other visuals, filters, or slicers in the report, providing relevant commentary based on the current data context.
• Users can customise the tone, terminology, and detail level in the narratives to match their preferences or organisational communication styles.

Cons:

• General observations and broad audience: It might not capture the specific context or nuances of your data. They work well for a broad audience but may not be tailored to the unique needs of a particular team or user.
• No priority: The tool cannot prioritise the insights it generates. This means crucial information or metrics requiring immediate attention might get buried amongst other findings, making it harder to focus on what's most important.
• Summary limits: There's a limit to the number of summaries that can be generated so Smart Narratives picks the most interesting things to summarise about the visual. It generates up to four summaries per visual and up to 16 per page.
• Compatibility: There are a lot of functionalities that smart narratives does not support including pinning to dashboard, publish to web, On-premises Analysis Services etc.

Latest AI Visual Tool

Copilot

Power BI Copilot is a new AI assistant built into Power BI that helps us to create reports and analyse our data faster and easier. Need to keep in mind that Copilot for Power BI is currently in preview phase, many of the announced features in Copilot are not available to the public yet.

Pre-conditions:

• Power BI Copilot applies to both Power BI service and desktop.
• To access Copilot in the Power BI service, the workspace must be running on F64 or Premium capacity.
• To use Copilot in Power BI Desktop, you need admin, member, or contributor access to at least a single workspace that is assigned to a paid Fabric capacity (F64 or higher) or Power BI Premium capacity (P1 or higher) that has Copilot enabled. That means Copilot for Power BI is (currently) better suited to larger organisations. This may change in future.
• Administrator needs to enable Copilot in Microsoft Fabric and allow Quick Measure Suggestions in preview features (normally Copilot capabilities in Microsoft Fabric are enabled by default in the Fabric admin portal).
• There are also some geographical limitations (outside US and France).

Capabilities:

Power BI Copilot offers a range of powerful features, with four functionalities likely to be the most attractive to users:

• Creating DAX formulas: Utilises Natural Language Processing (NLP) to create DAX measures. By providing clear field names, table names, and specific requests (e.g., YTD), it generates DAX formulas that can be used as measures.
• Auto-Creating reports: The most exciting capability of Copilot is to analyse existing Power BI datasets and suggests dashboard elements (even entire page) by inputting desired outcomes to chat pane. Users can then edit and refine these auto-generated reports by clicking the Edit button, explore different visualisations. This is game-changing capability to make report development much easier and faster.
• Generating Summaries: By inputting text requests in Copilot chat pane, it can generate easy to understand narrative summaries of semantic model, individual visual, specific pages, or even entire reports.
• Interactive Analysis: Users can interact with Copilot by asking questions or describing their needs, and Copilot provides the best answers. It can assist with analysis tasks like finding key influencers, identifying outliers, even creating forecasts.

Considerations:

As Copilot is still in its early-stage development, there are some important considerations to keep in mind: Accuracy Issues: Copilot may generate incorrect content if it misinterprets requests or cannot understand complex intents, leading to irrelevant results. It performs best when using simpler prompts with minimal conditions. The long prompt will still generate a report, it will be less accurate due to complexity of the prompt.

Limitations: There are constraints on visual modifications, layout changes, and adding filters/slicers. Licensing Costs: Copilot requires an expensive license.
Data privacy: users need to opt in to allow their data to move outside of their region to an Azure OpenAI endpoint in a different region. For data located outside the US or France, the Copilot preview is disabled by default unless the tenant administrator enables the setting to process data outside the tenant’s geographic region.

By understanding these capabilities and considerations, users can effectively leverage Power BI Copilot to enhance their data analysis and reporting processes.

To summarise

While Power BI's existing AI tools are powerful, Copilot introduces new possibilities for data analysis and reporting. Understanding the strengths and limitations of these AI tools is crucial to making informed decisions and ensuring accurate, reliable insights from your data.

However, as powerful as AI tools are, they can't completely replace human expertise - at least not yet. We still need to apply critical thinking, context, and our own knowledge on top of what the AI suggests.

By combining the best of AI capabilities with human skills, we can create incredible value from data. AI allows us to work smarter and faster, while human intuition ensures accurate, impactful insights. As AI in Power BI keeps evolving, we have a great opportunity to make data analysis more accessible to everyone, not just technical experts. But we must be responsible about implementing AI ethically and strategically.

Proper governance policies and best practices are crucial when introducing transformative technologies like AI into your data ecosystem. Our team at Ei Square can work closely with you to establish robust governance frameworks, ensuring data privacy, security, and ethical AI adoption aligned with your organisation's values and regulatory requirements. Get in touch for an initial discussion.