For years, the entry point to data has been the dashboard: a panel you open, filter and navigate until you find the figure you were after. But 2023 is bringing a deeper shift. In the wake of ChatGPT, the big BI vendors have started promising something different: that you write a question to your data, in your own language, and get an answer back. The technology is still emerging, but it is worth understanding now.
From the dashboard to the question
The dashboard does not disappear, but it stops being the only path. The idea behind natural language is that a business user — without knowing SQL or mastering the tool — can type “how much did we sell last quarter by region?” and get a table or a chart without depending on the IT team.
It is an appealing promise because it tackles the classic BI bottleneck: the distance between the person who has the question and the person who knows how to build the report.
What is natural language query (NLQ)
Natural language query (NLQ) is a platform’s ability to translate a question written in human language into a query against the data model, and to return the result as a figure, a table or a visualisation.
It is worth separating two generations that coexist today:
- Classic NLQ, based on recognising entities (fields, measures, filters) within the sentence and mapping them to the model. Reliable, but rigid: you have to ask the way the tool expects.
- Generative-AI-assisted NLQ, only just announced in 2023, which uses large language models to interpret more ambiguous questions and even generate explanations. Promising, but at very early stages.
How the platforms approach it in 2023
The three tools we work with come at natural language from different starting points:
- Power BI. Microsoft has long offered Q&A, its natural-language question box over a model. The big news this year is Copilot in Power BI, announced at Build on 23 May 2023 and released as a public preview: it promises to generate reports, DAX formulas and narrative summaries from natural-language instructions. Today, it is a preview.
- Tableau. Salesforce announced Tableau GPT and Tableau Pulse on 9 May 2023, its generative-AI capabilities built on Einstein GPT, promising automatic summaries and conversational exploration. At the time of writing they are in pilot, not general availability.
- Qlik. Qlik’s approach in 2023 is Insight Advisor, which combines its associative engine with augmented analytics to suggest analyses and answer questions. It is not a generative or agentic capability: it is augmented analytics over the associative model.
The common pattern is clear, but so is its maturity: these are announcements and previews, not settled products.
The key is the semantic layer
Here comes the part the headlines tend to skip. For the answer to “how much revenue did we make last quarter?” to be reliable, the platform needs to know exactly what “revenue” means in your organisation: which table, which filters, what is included and what is not.
That definition lives in the semantic layer and in well-defined metrics. Without them, natural language does not remove the old BI problem — two reports giving different figures for the same question — it accelerates it: now anyone can ask, and anyone can receive a plausible but wrong answer.
Put another way: the conversation is the easy part. The hard part, and the one that decides whether this works, is the clean data model and the agreed definitions underneath it.
What to expect
Natural language in BI is, in 2023, an emerging technology: plenty of promise, plenty of previews and little track record in production. Our recommendation is to treat it as what it is — a promising preview — and to prepare for it without yet depending on it.
The best investment today is not the fashionable feature but the unglamorous work that will make it useful: getting the data model in order, defining the metrics and governing who sees what. It is the same foundation that underpins the generative AI applied to Business Intelligence everyone is talking about this year.
At Digital Fox Data we help build that foundation and assess which Business Intelligence tools fit your organisation, so that when natural language matures, your data is ready to answer well. If you want to prepare the ground, let’s talk.