Power BI has a built in Q&A feature. It lets users type a question and receive a response drawn from the underlying data model. In principle, it is a useful idea. In practice, it often falls short of what business users actually need.
The built in Q&A visual interprets natural language questions and returns charts, tables or other visual elements as its answer. For many users, this creates friction rather than clarity. If someone wants to know the average cap rate for a particular neighbourhood, they do not want a bar chart. They want a number. If they are looking for the top five transactions by dollar amount, a ranked list with property names and values is far more useful than a column chart that requires further interpretation.
We encountered this gap while working on a commercial real estate analytics platform. The data was rich, the model was well structured, and the reports were performing exactly as intended. But when it came to quick, conversational queries, the native Q&A experience was not delivering the kind of answers that felt intuitive to the people using the reports day to day.
What we built instead
We developed a custom Power BI visual that works as a lightweight Q&A chatbot. Rather than accepting freeform text input, it presents users with a curated dropdown of predefined questions. When a question is selected, the visual computes the answer directly from the filtered dataset and returns it as clean, formatted text.
The layout is compact. A dark header bar sits at the top with a collapse and expand toggle, making it easy to tuck the visual into a corner of an existing report page without disrupting the overall design. Below the header, a dropdown lists the available questions. Below that, the answer appears immediately, styled with clear formatting, bold property names, properly formatted currency values and descriptive context.
The questions themselves are tailored to the specific dataset. For our commercial real estate platform, these include queries like the top five transactions by dollar amount, average cap rate, average price per square foot, average price per buildable square foot, the total number of transactions, and a breakdown of transaction volume by property type. Each question maps to a computation function within the visual that sorts, filters or aggregates the data and returns a concise textual answer.
Why this approach works better
The advantages over the native Q&A visual are significant, and they go beyond aesthetics.
Text answers instead of charts. The built in Q&A visual responds with visual elements. That makes sense for exploratory analysis, but for users who want a direct answer to a specific question, a chart introduces unnecessary steps. Our visual returns the answer as text. If the user asks for the average cap rate, they see the percentage and the number of transactions it is based on. Nothing more, nothing less.
Predictable, curated questions. Freeform natural language input sounds appealing, but in practice it introduces ambiguity. Users phrase questions differently, and the Q&A engine does not always interpret them correctly. Predefined questions remove that uncertainty entirely. Every question is mapped to a specific computation, and the answer is always accurate and consistent.
Full design control. The native Q&A visual offers limited styling options. Our custom visual gives us complete control over the layout, typography, colours and formatting. The dark header, the clean dropdown, the metric cards for single values, the numbered lists for ranked results. Every element is designed to match the report's visual language and feel native to the dashboard.
No external dependencies. The visual computes all answers locally from the data that Power BI passes into it. There are no API calls, no external services, and no internet dependency. This means zero latency, zero cost per query, and no risk of unexpected behaviour from an AI engine. The answers are deterministic. The same data with the same filters will always produce the same result.
Filter reactive. Like any well built Power BI visual, the chatbot responds to slicer and filter changes automatically. If a user selects a specific borough, date range or property type, the answers update immediately to reflect the filtered dataset. There is no need to reselect the question or refresh anything manually.
Works everywhere Power BI works. Because the visual is built as a native TypeScript custom visual using the Power BI Visuals SDK, it works across Power BI Desktop, the Power BI Service, Publish to Web, and embedded environments including Salesforce. There are no platform restrictions.
How it handles the data
One of the more important technical details is how the visual handles data at scale. Power BI custom visuals receive data through a dataView object, and by default, this is limited to a relatively small number of rows. For datasets with thousands of transactions, this can result in incomplete answers.
We addressed this by configuring the visual to request data in larger windows and to fetch additional segments when needed. The visual detects whether Power BI has more data available and requests it automatically until the full filtered dataset is loaded. This ensures that aggregations like averages and counts are always based on the complete dataset, not a truncated sample.
The visual also performs deduplication at the row level. In many real world data models, the same transaction can appear across multiple rows due to dimensional relationships. The visual groups rows by a unique transaction identifier and averages the numeric values, mirroring the logic typically handled by DAX measures in the report model.
A better fit for how people actually use reports
The reality is that most report users are not data analysts. They open a dashboard, apply a few filters, and want quick, clear answers. A compact Q&A chatbot with predefined questions fits that workflow perfectly. It sits quietly in the corner of the page, takes up minimal space, and delivers exactly what the user is looking for in plain language.
The question set is easy to adjust. Adding a new question is a matter of defining the computation logic and mapping it to the relevant fields. Removing or reordering questions is equally straightforward. This makes the visual adaptable as business needs evolve or as new metrics become relevant.
We built this initially for a commercial real estate analytics platform, but the pattern is entirely reusable. Any dataset with well defined KPIs and a clear set of questions that users ask repeatedly is a candidate for this approach. Retail, finance, healthcare, logistics. The architecture is the same. The questions and computations simply change to match the domain.
If your Power BI reports could benefit from a smarter, more user friendly way to surface key insights, we would be happy to discuss how a custom Q&A visual could work for your data. Get in touch with Plumgate to explore what is possible.