Forrester recently published our latest evaluation of business intelligence (BI) platform offerings: The Forrester Wave™: Business Intelligence Platforms, Q2 2025. There are two important takeaways from the research that are contrary to the media narrative on business intelligence:

  1. BI is alive and well. For the past 20 years, Forrester has consistently encountered predictions from other analysts and vendors claiming that “BI is dead.” These predictions have repeatedly been proven incorrect. BI continues to be a crucial enabler in the data-to-decisions process, which is essential for data-driven enterprises.
  2. GenAI is not the end of BI. Generative AI (genAI) is not replacing business intelligence; instead, it’s leveling the playing field as all BI vendors integrate generative and agentic AI capabilities based on the same large language models (LLMs).

Every BI vendor is now making a claim that their platform is genAI- and agentic AI-based. But how BI vendors are leveraging the power of genAI is what makes the difference. This means that data, analytics, and AI leaders must be very clear on how each BI vendor is delivering genAI-based functionality embedded in the BI platform. Here’s a small sample of genAI evaluation criteria that our research uses:

  • GenAI functionality. What specific BI platform functionality is based on genAI (natural language query [NLQ], natural language generation [NLG], enriching semantic layer, etc.)? Does the platform offer any non-core BI, genAI-based functionality such as mining unstructured data, documenting data sets/populating a data catalog, or others?
  • GenAI architecture. What specific foundation LLM APIs and/or BI vendor proprietary language models are provided out of the box? Can different LLMs be used for different genAI tasks (i.e., one for NLG, another for NLQ)? Can customers bring their own LLM license/keys?
  • GenAI domain specialization. Does the platform come with industry-/business domain-specific language models? Does the platform provide utilities for clients to fine-tune their own version of foundation LLMs and/or create custom expert language models?
  • GenAI for enterprise data. How does the platform’s genAI functionality get access to enterprise data, metadata, and other contexts? Is this capability based on retrieval-augmented generation (RAG) and/or other prompt engineering techniques? Is the capability based on fine-tuning LLMs and/or building expert/custom language models?
  • GenAI guardrails. What are the guardrails the platform uses to ensure that users only have access to LLM results they are allowed to see? Are the input guardrails solely based on prompt engineering and RAG or other techniques, as well? What are the guardrails the platform uses to ensure that LLM output complies with enterprise requirements such as content blocking/toxicity detection, content moderation, data/info validation, comparing design/runtime outputs, etc.?

For the detailed results, including all eight genAI criteria, 19 other criteria around vendors’ current offerings, and six vendor strategy criteria, please read The Forrester Wave™: Business Intelligence Platforms, Q2 2025, and/or set up a call with me.