If you've ever tried to build a production-grade AI agent, you've likely run into the same brick wall: the data. Not the model quality, not the prompts, not the infrastructure — the data pipeline. The lag, the copies, the sync jobs, the fragmented governance. For decades, enterprises have lived with the complexity of keeping operational databases and analytical systems in separate silos, patched together with ETL pipelines. For human analysts, that was a headache. For AI agents, it's a dealbreaker.

At the Data + AI Summit on June 16, 2026, Databricks made a bold claim: they've solved it. Here's everything you need to know.

The Root of the Problem: Why Data Pipelines Kill AI Agents

To understand why this announcement matters, you first need to understand why the old architecture is fundamentally incompatible with agentic AI.

For four decades, transactional and analytical workloads have lived in separate systems: operational databases served applications, while analytical systems answered questions.

That separation was manageable when queries were run by human analysts at human speed. But AI agents don't work on human timescales.

For humans, it is a headache. But for agents, it doesn't work. Agents are always-on, reasoning in loops, and their ability to act depends entirely on their ability to query complex enterprise data fast.

The old approach forced enterprises into a painful choice.

For years, enterprises that needed low latency at high concurrency had one option: stand up a separate real-time serving layer alongside the lakehouse. But that serving layer brought vendor lock-in, increased infrastructure costs, fragmented governance, and data that's never truly real time because it's always a copy. This left enterprises with a forced compromise: accept latency or fragment the stack.

The bet here is that AI agents, not people, will become the primary users of the enterprise data stack, and that the infrastructure beneath them has to be rebuilt for them.

What Databricks Actually Announced: Two Products, One Big Bet

At the Data + AI Summit, Databricks announced two products aimed at collapsing that infrastructure: Lakehouse//RT and LTAP (Lake Transactional/Analytical Processing).

Together, these two products take aim at opposite ends of the same problem — query latency and data unification — in a coordinated architectural bet.

As Databricks co-founder and CEO Ali Ghodsi put it:

"For decades, complicated data infrastructure was a tax that teams were forced to pay. Then agents arrived. In a matter of months, organizations effectively doubled their workforce, just not with humans. Agents write code, make calls, and run loops at a pace human teams never could."

Lakehouse//RT: Millisecond Queries Without Moving Your Data

The first product, Lakehouse//RT, directly targets query latency — the speed at which an agent can retrieve information to reason and act on.

Lakehouse//RT allows enterprises to run real-time analytics directly on governed Delta Lake and Apache Iceberg data, eliminating the need to set up separate serving systems to achieve millisecond performance.

The engine powering it is brand new.

At DAIS 2026, Databricks announced Lakehouse//RT, powered by Reyden, a ground-up engine rewrite and not an update to Photon. It runs against existing Delta and Iceberg tables without restructuring, requires Unity Catalog, and keeps data in open formats.

The performance numbers are significant.

Lakehouse//RT delivers sub-100ms latency at 12,000 queries per second, with response times as low as 10ms on smaller datasets and up to 16x better performance than existing dedicated serving stacks.

Crucially, governance travels with the data.

Every Lakehouse//RT query runs natively within Unity Catalog's governance framework with no separate permissions layer, no proprietary formats, and no sync/CDC pipelines, eliminating the cost and complexity of maintaining a separate real-time serving layer alongside the lakehouse.

Real-world adopters are already seeing benefits.

Companies like Magnite and Cisco are leveraging Lakehouse//RT to achieve consistent low-latency performance directly on governed lakehouse data, simplifying their pipelines and retiring separate serving systems.

LTAP: Eliminating ETL Pipelines at the Architectural Level

The second product, LTAP, attacks an even older problem — the ETL pipeline itself.

Databricks introduced Lake Transactional/Analytical Processing (LTAP), a new data processing architecture that unifies transactions, analytics, streaming, and operational data on a single copy of storage in the lake.

LTAP is Databricks' answer to the failure of HTAP (Hybrid Transactional/Analytical Processing) approaches. Instead of converging query engines, LTAP unifies data at the storage layer via the Lakebase architecture, a serverless PostgreSQL service.

LTAP stores Lakebase data directly in Unity Catalog using Delta and Iceberg formats, eliminating ETL pipelines, replicas, and synchronization overhead between operational and analytical systems.

Databricks claims to be the world's first LTAP platform. It combines Lakebase (serverless Postgres on open object storage) with the Lakehouse under a single governance model, source of truth, and storage layer for all operational, analytical, and streaming data.

This is no small claim.

If it delivers at scale, data teams building agent-first applications may no longer need to maintain distinct ETL pipelines, replica databases, and analytical lakehouses as separate infrastructure layers.

It's worth noting that

LTAP is listed as "coming soon" as part of Lakebase

, so full general availability is still on the horizon.

The Bigger Picture: An Agent-First Data Stack

Lakehouse//RT and LTAP don't exist in isolation. They're part of a broader platform rethink that Databricks is executing at pace.

Lakehouse//RT makes even more sense when viewed alongside Databricks' other announcements: Lakebase provides a PostgreSQL-compatible operational database, LTAP unifies transactional and analytical storage on a single copy of data, and Lakehouse//RT provides millisecond analytical serving directly from that same data. Together, Databricks is attempting to collapse what has historically been three separate systems.

Beyond infrastructure, Databricks also expanded its Agent Bricks platform.

Databricks expanded Agent Bricks into a comprehensive developer agent platform, and per the official Databricks blog, over 100,000 agents have been built on Agent Bricks since launch, with the platform processing more than 1 quadrillion tokens per year.

Complementing this is Genie One.

Databricks announced Genie One, an all-new agentic coworker that helps business teams automate and orchestrate their work across any data — structured or unstructured, analytical or operational, inside or outside Databricks.

Genie Ontology continuously extracts and updates business context from across Databricks and connected systems, so Genie One can look up the real answer to business questions in curated, authoritative data through SQL rather than reason from fragments spread across documents.

The broader market is also signalling the same consolidation direction.

According to VB Pulse Q1 2026, a three-wave longitudinal survey of 100-plus employee organisations, hybrid retrieval intent tripled from 10.3% to 33.3% across the quarter while standalone vector database adoption declined across every tracked vendor.

Practical Tips: How to Act on This Now

Whether you're a data engineer, AI architect, or technical decision-maker, here's how to respond to these announcements strategically:

Powered by Reyden, a new compute engine built for the concurrency and latency demands of modern agentic enterprises, Lakehouse//RT is now available in Beta

— sign up early to benchmark it against your existing stack.

LTAP eliminates pipelines synchronising operational and analytical stores, replicas to maintain, or connectors moving data between systems.

Quantify how much engineering time your team currently spends on this — it'll make the ROI case clearer.

Just provide a high-level description of the agent's task, and connect your enterprise data — Agent Bricks handles the rest. It is optimised for common industry use cases, including structured information extraction, reliable knowledge assistance, custom text transformation, and orchestrated multi-agent systems.

Customers who have built ETL pipelines and dedicated OLTP systems on top of Databricks face migration costs before LTAP delivers its promised simplification

— plan ahead rather than reacting at GA.

Conclusion: The Data Stack Is Being Rebuilt for Agents

Databricks' announcements at the Data + AI Summit 2026 represent more than product updates — they represent a fundamental rethinking of what enterprise data infrastructure should look like in an agentic world.

Databricks' proposal to eliminate copies and synchronisation between operational and analytical systems represents a significant step forward for agentic workloads.

The combination of millisecond query performance, unified storage, and governed access isn't a nice-to-have for AI agents — it's table stakes.

The question is no longer whether your data stack needs to change. It's whether you get ahead of the change or react to it. Now is the time to evaluate your current architecture, benchmark Lakehouse//RT against your serving layer, and start planning for an LTAP-native future.

Ready to future-proof your data infrastructure for the agentic era? Start by auditing your current pipeline complexity and signing up for the Lakehouse//RT beta today — your AI agents are only as fast as the data beneath them.