It’s been a year since generative AI (GenAI) burst into the public zeitgeist in the form of ChatGPT. Since then, the ecosystem has exploded, and numerous developments have appeared at a rapid pace. Above the surface, GenAI-powered agents and apps have emerged. Below the surface, other technology has emerged, including larger models, AI safety, and LLMOps. Gartner released its own Hype Cycle on the topic (spoiler alert: GenAI is peak hype), and with OpenAI’s recent release, there’s enterprise, consumer, and everything in between.
Analytics is an obvious use case of GenAI. Having a conversation with all your data and getting contextualized answers is a natural next step. But with so many vendors touting GenAI for BI, and analytics emerging from various backgrounds (e.g., BI, data science platforms, augmented analytics, decision intelligence, and more), it can be hard to keep up and discern what’s moving the analytics needle vs. what’s hyped.
Below is a roundup of GenAI offerings we’ve seen announced or launched in the BI and broader analytics market. Many companies have provided similar implementations of these transformative models, but there are some very unique applications as well. We aim to note where there’s something unique and what differentiates the implementation of the LLM from other analytics platforms.
In This Post
Aible is leveraging generative AI to create a new tool called ChatAible. ChatAible autogenerates an executive summary from an uploaded data set. The tool also provides industry-specific context for the uploaded data set via their GPT-4 integration. ChatAible double-checks the data to make sure the answer provided by GPT-4 is true. Users can ask follow-up questions of the analysis in natural language.
- What’s Interesting: ChatAible’s “executive summary” provides a narrative of the analysis in natural language, highlighting sections where it’s confident of the output.
- Our View: But what about the other text? If ChatAible isn’t confident enough to highlight the text, why is it providing you with this summation? This text may contain hallucinations, and the service’s lack of confidence should give executives pause about whether to implement the technology.
Amazon QuickSight Q
Amazon has brought generative AI to their QuickSight business intelligence platform with Q.
Amazon QuickSight Q brings natural language search to the platform, enabling users to ask business questions and instantly create visualizations. Q will also enrich metadata with automated synonym generation to increase the accuracy of search. It will also enrich the semantic layer by automatically tying things like place names to a location. Finally, this new capability adds key driver analysis, forecasting, and a narrative explanation of the causes of change to Amazon QuickSight.
- What’s Interesting: Amazon QuickSight packs a lot of features into its natural language search capabilities with Q. Its automated insight generation in natural language search is not a common feature we’ve seen across the generative AI implementations featured.
- Our View: Amazon QuickSight Q’s implementation of natural language search and automated insights is still new. Other platforms have more robust offerings around automated insights and do not need to rely on LLMs for insight generation. This removes risks inherent with LLMs.
AnswerRocket, an analytics platform focused on CPG solutions, has created a chatbot named Max, powered by GPT-4, to assist with search and analysis. Max enables users to ask questions in natural language and receive a response in natural language while also allowing for follow-up questions (like ChatGPT). Max can also provide detailed narratives and visualizations to accompany answers. Max provides some data preparation features, including automated data classification, definitions, synonyms, and suggested questions.
- What’s Interesting: AnswerRocket’s Max was one of the first chatbots revealed in the flood of chatbot announcements this past spring.
- Our View: Many products have come along that provide similar functionality on more developed platforms.
ChatGPT Code Interpreter Plug-In
The ChatGPT Code Interpreter plug-in provides ChatGPT with the ability to run data analysis and create visualizations by uploading files for analysis (CSV, Excel spreadsheets, or even SQL databases). You can then use natural language prompts to run analyses and create visualizations. It can also create Python code from natural language. Each session is also ephemeral, meaning when you close the session, the data provided and analysis will disappear. There’s also a limit to the amount of data allowed to be uploaded to the service.
- What’s Interesting: As a plug-in for ChatGPT, the ability to upload simple spreadsheets and run analysis by using prompts is a truly innovative idea. From a workflow perspective, this seems like one of the most intuitive ways to get started on an analytics platform.
- Our View: Being a plug-in for ChatGPT has its limitations: namely, a lack of data governance, which should effectively bar almost all large enterprises from using this type of service. The ephemeral nature of the tool will effectively restrict a number of use cases. Additionally, data volume cap limits this tool’s usefulness to very simple applications.
Dataiku, a data science-focused analytics platform, launched one of the first LLM ops platforms available on the market today. The Dataiku LLM Mesh provides a way for large enterprises to enable governance and deployment of various LLMs with integrations to the largest LLMs available today, including OpenAI’s GPT, Huggingface, Google’s LLMs, Cohere, and more. The platform provides visual LLM development and training capabilities. The suite of tools for LLM ops includes a PII detector, detailed cost estimates, the ability to cache responses to common questions, and audit logs. They’ve also included a chatbot to assist with the development of Python code.
- What’s Interesting: The concept of an LLM ops platform may be in its early stages, but the velocity of LLM development may put Dataiku ahead of the curve for the next decade.
- Our View: The ability to manage multiple LLMs and control governance should provide large enterprises with additional peace of mind when leveraging this new technology. Dataiku’s LLM Mesh should only be considered for very large enterprises and those that have a particular LLM-focused niche.
Google Looker + Duet AI
Google’s Looker business intelligence platform is benefiting from Google innovation on generative AI with the incorporation of Duet AI. The Duet AI integration brings conversational analytics to Looker as well as the ability to create reports/visualizations with simple natural language prompts. Duet AI also enables users to generate Google Slides presentations based on reports. Finally, with the Duet integration, Looker users will be able to create formulas and generate LookML code using natural language.
- What’s Interesting: Looker + Duet AI is taking generative capabilities to the next level by incorporating into Looker not just report generation but also Google Slides generation. This capability will further extend the ability for non-technical users to create presentations that drive data-driven decision-making.
- Our View: As a leader in the generative AI space, Google has begun to incorporate these key capabilities into their analytics platform. By adding AI-generated Google Slides presentations, they may be further democratizing data-driven decision-making. However, users should be cautious about implementing such features with current issues around generative AI (e.g., hallucinations, governance, etc.).
GPTExcel is a tool used to help streamline spreadsheet processes for Excel, Google Sheets, and Airtable. GPTExcel provides a natural language interface to generate formulas and create complex calculations using natural language prompts. It also includes natural language-to-SQL code generation, as well as the ability to leverage AI to understand regular expression patterns for data validation and filtering.
- What’s Interesting: Productivity gains across multiple platforms may be worthwhile for small use cases.
- Our View: Many analytics platforms are already providing this feature, and you can generate formulas and complex calculations in ChatGPT itself.
GPT for Sheets and Docs
GPT for Sheets and Docs is a Google Workspace plug-in enabling users to leverage GPT-3.5, GPT-4, and Anthropic’s Claude directly in Google Sheets and Docs. The plug-in enables users to run rudimentary data cleansing tasks directly in Google Sheets. Some of the features include being able to split text semantically by a number of parameters, summarize the content of your spreadsheet into natural language, and more. The plug-in requires an API token from either OpenAI or Anthropic. The plug-in is currently free in its beta period.
- What’s Interesting: The ability to choose your desired LLM, including models from OpenAI and Anthropic, provides a crucial distinction from many products available on the market today. In addition, data cleansing leveraging an LLM is an uncommon use case in the market today.
- Our View: Limiting this data cleansing to Google Sheets limits the general use of this tool for many organizations working with data at scale.
Hex, an analytics company, announced Hex Magic toward the start of 2023 to incorporate AI in their platform. Like other companies, Hex Magic starts as a conversational analytics tool. Ask questions, and Hex Magic provides answers. This generative AI capability also enables users to generate, validate, fix, and document SQL or Python code. Hex Magic maintains detailed knowledge of the data’s schemas and past operations to make better recommendations. Finally, this new AI integration will generate visualizations from natural language prompts.
- What’s Interesting: Hex Magic’s knowledge of a user’s database promises better answers than many of the implementations covered here.
- Our View: Hex Magic provides many of the core capabilities available in most of the generative AI offerings available today. It’s yet to be seen if the detailed knowledge of your data provides a meaningful improvement over other offerings.
Microsoft’s Copilot is the integration of GPT-4 with Microsoft’s 365 ecosystem, including Word, Excel, Teams, Outlook, Power BI, and more. Copilot in Power BI, Microsoft’s business intelligence tool, includes the ability to create reports and dashboards from natural language, as well as create simple analytical narratives around the data. Copilot in Power BI also has a built-in chatbot interface for conversational analytics.
- What’s Interesting: With a large ownership stake in OpenAI, Microsoft has access to some of the most sophisticated LLMs available today in GPT. It also has a large stack of platforms with various implementations of LLM technology. In analytics terms, Copilot in Power BI’s ability to generate a report from natural language is intriguing from a data democratization perspective.
- Our View: Being tied to OpenAI also has its negatives. We have yet to see a “bring your own model” functionality in Microsoft’s platform. We have also yet to see the cost implications of effectively leveraging Copilot in the Microsoft ecosystem.
MicroStrategy has introduced their new MicroStrategy AI tool, which has a varied feature set. The tool includes SQL generation from natural language, prompt-based narrative explanations of analysis, and reports generated directly from prompts. In addition, MicroStrategy’s LLM integration includes proactively suggested questions to help guide the start of your analysis. Finally, Auto Expert provides MicroStrategy customers with the ability to search through documentation and learning materials.
- What’s Interesting: Auto Expert, a unique spin on search for documentation, may lessen the headaches caused by working with MicroStrategy’s platform.
- Our View: Much of the functionality in MicroStrategy’s AI toolset is similar to that of other vendors in the market today.
Narrative BI is an analytics start-up focused on simplifying business intelligence with generative AI capabilities. The tool allows you to connect your data from Google Analytics, Google Ads, Facebook Ads, and Hubspot to derive narrative insights from the data. These narratives are generated on a monthly or weekly basis and leverage GPT for the generation.
- What’s Interesting: From a workflow perspective, this tool may have the potential to drive adoption with a very intuitive interface.
- Our View: The integrations available limit the number of potential use case applications for this tool.
Oracle Analytics has released a number of generative AI-powered features in the past year. AI Auto-Insights automatically generates visualizations from datasets. Machine learning can be used to automatically explain a metric or an attribute. Other simplified ML-generated insights include clustering, forecasting, and ARIMA-based algorithms.
Oracle is experimenting with other integrations, including AI vision, for object detection and image classification, sentiment analysis for text, and document understanding to extract values from receipts, resumes, and more. Oracle has also partnered with Synthesia to bring AI avatars to analytics presentations, which provides a human face and gen AI-based audio to explain analysis.
- What’s Interesting: Oracle’s Auto-Insights and simplified ML tools bring them closer to industry leaders like Tellius in their automated insight generation capabilities. The investment in AI experiments shows promise and could unlock new use cases in the future.
- Our View: Oracle’s automated insight generation is a step toward democratizing data access in organizations. Oracle’s experiments in analytics seem to be wide-ranging, and the intended purpose of these new capabilities is unclear today.
Pyramid Analytics, a business intelligence platform, has a diverse array of GPT-enabled features. The offering includes the ability to generate Python, R, SQL, DAX, and MDX from natural language, which is a wider selection of programming languages compared with other vendors on this round-up. Natural language-generated spreadsheet formulas are another option. In addition, Pyramid Analytics enables users to retrieve data from publicly available sources via prompt engineering. Finally, users can generate designs for content and graphics using the GPT integration.
- What’s Interesting: Pyramid Analytics’ capability to retrieve data from publicly available sources is a unique application of LLMs.
- Our View: LLMs’ knowledge base is limited to the date it finished training. For example, GPT-4 is limited to knowledge from prior to April 2023. This may limit the platform’s ability to retrieve useful, up-to-date information.
Sisense, a business intelligence company focused on embedded use cases, recently announced a plethora of features taking advantage of generative AI. The integration within Sisense Fusion enables users to create widgets leveraging the power of LLMs. One widget example enabling data enrichment includes the ability to append industries onto a list of companies. In addition, Sisense launched an analytics chatbot with suggested starting questions, as well as narrative summaries of analyses. Sisense has also designed their platform to be LLM-agnostic.
- What’s Interesting: Sisense continues to invest in their embedded offering with these features targeting analytics developers. In addition, not many platforms featured here have been designed to be LLM-agnostic, which should provide some future flexibility to Sisense users.
- Our View: Data enrichment leveraging an LLM seems like a risky proposition when hallucinations are still very prevalent from the largest LLMs today.
Tableau has brought generative AI to their business intelligence tool via a number of offerings. Tableau Einstein is a conversational analytics tool for data preparation and analysis. Einstein comes with a chatbot interface and includes the ability to ask questions in natural language. It also provides recommended insights to start the conversation, as well as the ability to perform simple data preparation tasks. Tableau also developed Pulse, a Slack integration for business users. Pulse provides narrative insights to users via Slack and allows users to dive deeper into the data.
- What’s Interesting: The breadth of features in Tableau’s application of LLM technology is quite impressive.
- Our View: There’s not much differentiation in Tableau’s latest AI-enabled offering. Most of these applications are available on other platforms.
Tellius, an augmented analytics platform, focuses on multi-persona flexibility and collaboration. Tellius, one of the industry leaders in natural language analytics, has been focused on innovating natural language processing and automated insight generation technology since 2017. Tellius announced its LLM-based offering, Copilot, in spring 2023. Copilot offers greater context and narrative summaries for business users, metadata enrichment for analysts, and automated code generation/validation for data science teams. Tellius is also incorporating LLMs into its natural language search to help improve business context understanding. Automated code generation provides prompt-based Python and SQL code. Automated metadata generation helps provide synonyms from column headers to enrich search.
- What’s Interesting: Tellius, one of the earliest analytics platforms to provide natural language search, doesn’t rely on LLM technology as the backbone of its computational engine. Rather, Tellius views GenAI as a means of enriching search with LLM-based contextual understanding for a given industry or use case. This will help to increase the accuracy of search results for a given prompt, which essentially means there’s no chance of hallucinations. Tellius also includes explainable AI features to help understand how an analysis was derived. AI explainability is unavailable on most LLM platforms today.
- Our View: The application of LLM technology on a platform built for multi-persona analytics (i.e., a platform for data scientists, analysts, and business users) provides more opportunities for collaboration among groups. An intuitive platform providing automated actionable insights also helps an entire organization become more data-driven.
ThoughtSpot recently unveiled their own GPT-4 powered Sage functionality to enhance natural language search, suggestions, and data modeling. Sage has replaced ThoughtSpot’s search engine with one powered by GPT-4. This new search allows users to provide human-in-the-loop feedback on search results. Sage provides an interface to ask questions in natural language and receive visualizations as answers along with a short narrative analysis. ThoughtSpot Sage can also automatically generate synonyms for metadata.
- What’s Interesting: ThoughtSpot Sage has many of the features included in other analytics platforms and may have the potential to increase accuracy in search results.
- Our View: ThoughtSpot has had limited success in developing a true natural language search platform. It’s unclear if their attempts to leverage LLMs to enrich search have provided more accurate results.
Generative AI will be an industry-changing technology in the near future. However, many companies’ current implementations of the technology are similar. Almost everyone on this list, for example, has created an LLM-based chatbot as an interface for search and narrative summaries. The focus on conversational analytics makes sense for the initial development, as we’ve all played with ChatGPT at this point. It’s the use case with which the company that makes the world’s most popular LLM, OpenAI’s GPT, launched to the world!
Conversational analytics is absolutely a great use case for LLMs, but we’re curious if the LLM-powered chatbot interface will be a success or a dud in democratizing data access. There are some helpful features provided by some of the companies on the list, including suggested queries, but there’s still the question of how a business user will be able to ask the right questions of the data and get a successful output. A business user needs to be able to vet if the information provided is correct, or the proposition of LLM-enabled self-service analytics falls on its face.
Generative AI is a brand new canvas on which development is just getting started. For LLMs to succeed, analytics companies need to start rethinking how they approach the data analysis problem. It’s hard to think about all the potential use cases of the technology, but there’s a lot of potential for start-ups to kick-start their innovation with LLMs.
There’s so much inertia to development and adoption, especially at large enterprises, that it’s hard to imagine the large analytics providers at the forefront of development for very long. We recommend researching and understanding what leaders in the LLM field are working on and exploring potential use cases at your organization.
LLMs are here to stay and will only become more powerful in the near-term future. With the pace of innovation, if you’re not keeping up with new developments, you may be left behind. We wait with great interest to see how LLMs are enabling better analytics experiences in the future.
If you would like to learn more about Tellius Copilot, you can request access here.