What is the modern data experience?

Business is won or lost based on the quality of the experience you deliver to customers, partners, vendors, and employees. These experiences are built entirely on data. Harnessing data to deliver value is the single most powerful way to engage today’s demanding consumers—not to mention capturing market share and accelerating strategic decision-making. 

But there's a problem. 

The language barrier between humans and machines makes it impossible for many of us to access data insights and analytics. Since we couldn’t teach machines to speak and think the way we do, we had to learn to speak the way they do—in code. If you don’t “speak data,” you are often forced to rely on those who do—the analysts and data scientists who can translate the data into actionable insights. That means an unending queue of requests for data teams, while business people are often relying on gut instinct to make decisions.

That’s been the only option—until now. 

Bridging the human-machine language gap 

With the arrival of GPT and other large language models (LLM), machines have learned to speak our language. It’s hard to overstate how big of a deal this is. Foundational language models have changed everything about how we interact with machines.  

With the modern data experience, you can query data the way you’d ask your colleague a question. You can use natural language, in all its glorious flexibility, imprecision, and vagueness, to get an accurate answer. At ThoughtSpot, we’ve spent the last decade building a platform that empowers anyone to talk to the data and ask their own questions in natural language. But now, with our integration with GPT, the data can talk back. 

The relationship between data and search

At ThoughtSpot, it’s always been our mission to build data analytics systems and UX that are orders of magnitude easier to use than anything else on the market—without compromising security, accuracy, or trust. We want to empower everyone with self-service analytics to search their own data, finding the answers and insights they need without learning how to code. 

To take on this massive task, we split the problem in two: 

  1. Make a backend system that can deduce user intent from natural language queries—and then use AI to model the semantic layer, fetch the right data, compute the values, formulate the precise answer, and create interactive visualizations of the answer.

  2. Build a natural language frontend to distill the user intent from the question, prompting the backend to get the job done, and then provide a natural-language answer to the user.

Today, our backend is right there—handling complex data, sitting on every cloud data platform, integrating with all of your favorite tools, and performing at the scale of trillions of rows at lightning speed. But in the second category—the front end—the language models we needed simply didn’t exist yet. 

So, we focused on building the secure, controllable infrastructure necessary to make natural language work with data—and watched foundational language models mature until they were truly ready.

Which they eventually became. Right about…now. 

How GPT and AI are changing search

Foundational language models are powerful tools that are creating a new relationship between humans and machines. However, to truly transform the way we search for answers in a business context, the technology needs to be coupled with AI. 

If you’ve played with GPT, you’ll already know that it has limitations—it can confidently answer any question you put in front of it, but you can’t trust all of those answers. GPT still stumbles when it interprets complex data, especially math tables and calculations. This alone should give users pause when making critical business decisions.

But where things start to get exciting is—of course—search. The way most people access, consume, and utilize data will change dramatically thanks to GPT. With LLMs, we now have the ability to interact with data by asking simple, straightforward business questions. This capability paired with search will usher in a new wave of personalized data discovery.

That’s where ThoughtSpot and the modern data experience enter the picture. 

How ThoughtSpot is delivering the modern data experience 

While the teams at companies like Google and Microsoft have been building and training language models, we’ve been building the architecture that will make those models play nicely with data. We call it ThoughtSpot Sage, and it’s going to totally change how you think about data experiences. 

Here’s how it works 

Diagram showing the different processes for GPT-3 vs GPT-3 plus ThoughtSpot.
  1. When you ask a question, ThoughtSpot Sage starts by figuring out what data is relevant to the answer. Our system uses that analysis to limit what data GPT (or any other foundational model) should use to understand and answer the question. This makes it easier for GPT to parse the natural language question with a far greater degree of accuracy. 

  2. Then, Sage converts your question into a prompt for GPT that provides the relevant context it needs to be truly accurate. 

  3. Once GPT has parsed the sentence, it gives us abstract SQL on top of an abstract table. Sage then turns that abstract SQL into Keyword Tokens, and then those Keyword Tokens into a far more robust SQL that works with real-world, complex schemas and is optimized for your specific cloud platform. 

  4. Sage then uses AI to give you a natural-language explanation along with the answer. And, to make sure the answer is correct, it’ll also do one final check over the generated query to confirm. 

  5. In just a few moments, you’ll end up with an interactive chart with the answer to your question and a narrative to explain it.  

The business benefits of ThoughtSpot Sage

Look, we’re as excited about tech innovation as the next hard-core data geeks—but we’re more excited by the incredible business potential that happens when you integrate language models with AI-Powered Analytics. 

ThoughtSpot Sage lets you: 

  • Ask questions like a business leader—but get insights like a data analyst 

  • Tap the institutional knowledge of your business, whether by capturing it in the data model or by searching an analytics catalog

  • Uncover new insights with AI-informed data search and natural language explanations 

  • Scale your data’s impact with confidence and continue providing feedback to the model to improve your business results over time 

Intrigued? You can see the Modern Data Experience come to life by watching the sessions from Beyond 2023 on demand.