MicroStrategy update aims to improve generative AI accuracy

Generative AI has been the dominant trend in data management and analysis in the two years since OpenAI launched ChatGPT, which brought a significant improvement in the capabilities of large language models.

Two of the biggest potential benefits of generative AI are true natural language processing that allows workers of any skill level to query and analyze data to inform decisions and process automation that enables experts in data to be more effective.

As a result, vendors such as MicroStrategy, Domo And Qlik — among others — have created suites that allow customers to easily combine generative AI capabilities with their own proprietary data to develop tools that understand their business.

Accuracy, however, remains an issue for generative AI tools. Even when trained with quality data, generative AI models and applications sometimes provide incorrect or even bizarre responses called AI hallucinations.

Data quality helps reduce the risk of hallucinations. The same goes for the volume of data: the more quality data a model or applications are trained on, the less likely they are to constitute an answer. But they’re not foolproof, which is part of the reason MicroStrategy has provided explanations showing how Auto arrived at an answer.

Now, by upgrading the large language model that underpins Auto AI to OpenAI GPT-4oMicroStrategy hopes to further improve the accuracy of generative AI responses to make them more trustworthy. With GPT-4o working in concert with a company’s proprietary data, Auto now better understands the context of a user’s query, resulting in a more relevant and accurate response.

Beyond attempting to improve Auto’s response accuracy, new MicroStrategy AI features include Auto for Teams, an integration that allows users to query Auto directly within Microsoft Teams so they don’t need to leave their collaboration environment to work with data.

The suite also includes Auto Narratives, a tool that provides automatically generated text summaries and visualizations to provide users with context related to the data they are using. It also has a feature that allows developers to train bots using multiple data sets without needing to first consolidate their data into a single unified set.

Each is important and will benefit MicroStrategy’s customers, according to Leone.

“Auto for Teams is a feature that I think anyone with Teams would appreciate: just send a message to an expert on a certain topic and get a relevant response,” he said. “And… Auto Narratives really fuels the idea that explainability matters. Enabling more people to understand data remains a critical area for businesses, and this is a reliable shortcut.”

As MicroStrategy adds new features to MicroStrategy AI and attempts to improve existing ones, customer feedback is the primary motivation for deciding what to feature in a given product. platform updateaccording to PeggySue Werthessen, the vendor’s vice president of go-to-market strategy.

“The impetus for these enhanced Auto and actionable analytics capabilities was to work directly with our customers to understand how GenAI can deliver real business value,” she said.

Customer feedback is complemented by MicroStrategy’s ongoing efforts to ensure its platform is “accessible, flexible, and operates with reliability and scalability to support demanding use cases,” Werthessen added.

Beyond the new features in MicroStrategy AI, new features in MicroStrategy One include the following:

• New integrations with Microsoft Office 365 that allow users to import entire dashboards and other data assets into Excel and PowerPoint presentations, then keep these data assets up to date with automatic updates.
• A connector to Tableau allowing joint customers to access MicroStrategy datasets in their Tableau environments.
• Python Action Buttons, a feature that allows users to automate Python workflow in third-party applications like Salesforce and Workday from their MicroStrategy dashboards

Of note are the Python action buttons, according to Werthessen.

“This means that a front-line knowledge worker can launch workflows for individual items on a list with the click of a button,” she said.