Forrester Just Re-Graded 14 Enterprise Data Platforms. The New Test Is AI Agents.

A data lakehouse is a category of enterprise database that combines warehouse-style structured query with lake-style flexibility for raw and unstructured data, and it has become the default data layer underneath most large companies' analytics and machine learning. Forrester, an independent research firm whose periodic vendor evaluations (called Waves) are widely used by enterprise buyers, just re-scored 14 of them in its Q3 2026 evaluation, and the new test is no longer how well they store and query data. It is whether they can continuously serve trusted, governed, real-time context to AI agents that are actually executing work, not just answering questions in dashboards.

Forrester spelled out the shift in three takeaways from the Wave. The lakehouse is becoming an execution layer for agentic AI, meaning agents that take actions on data, not just retrieve it. Trust is the foundation of an AI-ready lakehouse. Open architectures are critical for long-term AI success. The criteria have been moving in that direction for two Wave cycles, but Q3 2026 is the iteration in which the weighting is explicit enough that vendors built primarily on storage and query benchmarks will see their scores move.

The new grading can be read as three concrete questions to put to any lakehouse vendor, and the vendors that have gone public about their own positioning give a sense of which way the field is heading.

Can the platform run agents, not just serve them data? Forrester's framing treats the lakehouse as an execution surface, not a passive repository. The vendor pitch has largely caught up: Cloudera, named a Leader in the Q3 2026 evaluation, has been public about repositioning its platform around the agentic AI era, and Databricks has introduced CustomerLake, an agentic customer data platform embedded directly in its lakehouse, rather than bolted on as a separate product. The Leader and one of the major challengers are both reorganizing around the same axis, which is a signal that the category is moving in one direction, not two.

A lakehouse that ships data to a model faster than it ships lineage, access policy, and freshness metadata is now a liability, because an agent acting on ungrounded context can take a destructive action in production. The Forrester recap flags governance, lineage, and real-time context as the gaps that most often separate the leaders from the rest. For a buyer, the test is whether the platform can answer, for any record the agent touches, where it came from, who is allowed to see it, and how fresh it is, without a separate data team in the loop.

Openness in the Wave context means the platform does not lock data, models, or workflows into a proprietary format. Independent buyer guides have been pushing on the same axis. The reason it is weighted so heavily now is that an AI agent stack is harder to migrate than a dashboard stack, because the agent depends on the platform's representation of context. Lock-in at the lakehouse level becomes lock-in at the agent level, and the cost of switching rises sharply.

Forrester's Q3 2026 Wave is the iteration in which the firm published a lakehouse evaluation with AI readiness, trust, and openness weighted explicitly alongside the old storage and query benchmarks, and the full Wave report carries the detailed scoring. For anyone choosing or modernizing a lakehouse now, the useful question is not who is on top in this Wave. It is whether your shortlist can answer the three new questions without a custom integration project.