Databricks Data + AI Summit 2026: Key Announcements Explained (Genie One, Genie Ontology, Databricks Lakebase, Unity AI Gateway, LTAP, AI Agents, Agent Bricks, Opensharing, and a lot more)

Databricks Data + AI Summit 2026 introduced one of the largest sets of platform updates in the company’s history. While many announcements focused on new products and features, a broader pattern emerged across the keynote sessions and technical tracks.

Several long-standing challenges that have limited enterprise AI adoption are now becoming major product priorities:

• Providing business context to AI systems
• Governing AI assets at scale
• Reducing operational complexity
• Bringing transactional and analytical workloads closer together
• Making enterprise data more accessible to business users

Taken together, the announcements provide useful insight into where the Databricks platform is evolving and which architectural trends enterprise organizations should monitor over the next several years.

This guide summarizes the most significant announcements and examines their implications from a technology and operating model perspective.

One Platform for the Agentic Era

Before diving into individual announcements, it is worth naming the single thread that ran through everything. Databricks is no longer positioning itself solely as a lakehouse or analytics company. The 2026 summit was a declaration that the lakehouse, the semantic layer, the agent runtime, and the governance layer are now one unified platform, built specifically to run AI agents safely at enterprise scale.

“Most enterprise AI today is just guessing with false confidence. That is not good enough for business. If you’re a CFO and AI can’t tell you why margins changed, or you’re a sales leader, and it can’t find your next upsell, that’s not an AI problem, that’s a context problem.”

— Ali Ghodsi, Co-founder and CEO, Databricks

Every product launched at DAIS 2026 is, in some way, an answer to that context problem: how do you give AI real, authoritative, live knowledge of your specific business, and then govern, monitor, and operate it at scale?

Genie One: Your Company’s AI Coworker

The most visible announcement at this year’s summit was Genie One.

Databricks positions Genie One as 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.

At first glance, Genie One may appear similar to other enterprise AI assistants currently available on the market. However, its differentiator is the degree to which it leverages existing Databricks assets, including Unity Catalog governance, enterprise data, business metrics, and platform-native AI capabilities.

The objective is not simply to provide conversational access to information, but to reduce the dependency on traditional analytics workflows.

What Genie One Does

Unlike earlier generations of enterprise AI assistants siloed by department or data domain, Genie One is designed as a universal business coworker. It can:

• Answer natural-language questions using governed enterprise data, pulling from SQL tables rather than guessing from fragments across documents
• Connect to Gmail, Slack, Teams, and other workplace tools to pull context from calendars, emails, and meetings
• Draft documents, generate reports, and schedule recurring tasks autonomously
• Explain why margins changed, surface upsell opportunities in a sales pipeline, or help finance close the books, grounded in the same operational data the business actually runs on

Pricing note: Genie One is enterprise-only (requires Unity Catalog governance). Each identified user receives 150 DBUs of free LLM usage per month, equivalent to roughly $10.50 at US East rates, or approximately 80-100 Genie questions per user per month. Usage beyond the free tier is billed pay-as-you-go starting July 6, 2026. There are no seat-based fees.

How It Compares to the Previous Genie

Genie began as a conversational analytics assistant inside Databricks AI/BI. Genie One is the next step: a data-smart AI coworker that extends beyond domain-specific queries to full agentic cowork, including schedules, alerts, monitoring, document creation, custom skills, and custom MCP server support.

Why does it matter for customers?

The classic bottleneck in most companies goes like this: a business leader has a question, sends it to an analyst, the analyst figures out what is actually being asked, pulls the data, builds something in Tableau or Power BI, and sends it back two days later. Genie One short-circuits that loop. Business users get answers grounded in real data without waiting in queue.

The accuracy numbers are striking. On a Databricks-designed benchmark of 28 real-world enterprise data questions, Genie answered 84.5% correctly on the first attempt. The best general-purpose coding agent on the same test scored 52.4%. The worst scored 25%. Genie also ran at 2× the speed of the strongest competitor.

Why does it matter for Databricks developers?

For engineers and data practitioners building on Databricks, Genie One is not just a business user tool, it is a surface for your own work. Genie Code (the developer-facing counterpart, also generally available) can write, debug, and explain pipelines, notebooks, and SQL. Beyond that, Genie One plugs into custom MCP servers, so anything you have built with the Model Context Protocol becomes Genie-accessible without any extra work.

The bigger unlock is Genie Agents and Genie App Builder (covered below), which let developers publish reusable business agents and full applications built on top of Genie’s data-grounded reasoning.

Genie Code also received an expanded full-page workspace at DAIS 2026, designed for longer-running, multi-step data and ML work across notebooks, files, and pipelines in a single interface.

Where to read more

Databricks Blog: Genie One, Genie Agents, and Genie Ontology: databricks.com/blog/introducing-genie-one-genie-ontology-and-genie-agents

Official press release: databricks.com/company/newsroom/press-releases/databricks-launches-genie-one-all-new-agentic-coworker-every-team

Genie Ontology: Teaching AI What Your Business Actually Means

The technical engine behind Genie One is Genie Ontology, arguably the most strategically important announcement of the entire summit.

The Problem It Solves

The reason most enterprise AI fails to answer real business questions is not model quality, it is context. Business context is scattered across dashboards, queries, pipelines, wikis, tickets, and chat threads. When AI cannot find this context, it fills gaps with inference, producing generic or wrong answers. Previous approaches tried to solve this with retrieval-augmented generation (RAG) over documents. Genie Ontology takes a fundamentally different approach.

What Genie Ontology Is

Genie Ontology is a live context layer that continuously learns your business from internal and external data, AI tools, and popular workplace apps. It automatically extracts snippets of knowledge from tables, queries, dashboards, pipelines, and connected apps, and organizes that knowledge into a living graph of how a company works and what its data actually means. 

This includes:

• Metric definitions — what ARR, NRR, or active user means in your specific business
• Business terms — unique calculations, departmental vocabulary, and domain-specific logic
• Relationships — how concepts, metrics, tables, and teams relate to one another
• Authority signals — using an approach similar to PageRank, Genie Ontology weighs definitions by source authority, frequency of use, ties to certified assets, and freshness

Why It Self-Improves

Unlike static semantic layers or manually curated ontologies, Genie Ontology continuously updates as new data, queries, and business changes occur. It feeds directly into Unity Catalog’s new semantic capabilities  (Business Glossary, Domains, and Metrics) meaning the better your semantic modeling, the smarter Genie becomes.

Why does it matter for customers?

Here is the honest truth about most enterprise AI pilots: they fail not because the model is bad, but because the model does not know your business. It does not know that your ‘active user’ definition changed in Q3 2024. It does not know that EMEA reports revenue in euros but the board deck converts to USD at month-end rate. It does not know that the sales team uses ‘closed-won’ to mean something slightly different than what is in Salesforce.

Genie Ontology captures all of that automatically, and keeps it current. The payoff is that every agent, every dashboard, every AI-generated report pulls from a shared, authoritative understanding of your business rather than each making its own assumptions.

Why does it matter for Databricks developers?

If you are building agents on Agent Bricks (see below), Genie Ontology is what gives those agents instant business awareness. You no longer need to hard-code business logic into every agent’s system prompt. Ontology acts as a shared context store across all your agents.

Practically, this means less prompt engineering, fewer hallucinations on business-specific questions, and agents that stay accurate as your business changes  without you having to update them manually. The ontology feeds directly into Unity Catalog’s Business Glossary, Domains, and Metrics, so the investment you make in one place pays off everywhere.

Where to read more

Databricks Blog: Genie One, Genie Agents, and Genie Ontology: databricks.com/blog/introducing-genie-one-genie-ontology-and-genie-agents

Genie Agents & Genie App Builder: From Conversation to Product

What Genie Agents & Genie App Builder are?

Genie Agents lets any business user turn a Genie conversation into a reusable, shareable agent that other people can access in Slack, Teams, or any connected surface. No engineering required. Genie App Builder goes one step further: describe what you want an application to do in plain language, and it generates a working web application backed by your lakehouse data, complete with authentication, permissions, and cost controls.

Under the hood, Genie App Builder is built on AppKit, a TypeScript SDK purpose-built for production Databricks applications that handles caching, telemetry, retry logic, and seamless data integration out of the box. Genie App Builder has direct awareness of your workspace: the tables you have, the Unity Catalog semantic layer you have defined, and the governance policies already in place.

Why does it matter for customers?

Think about the person in your organization who knows the most about, say, how to interpret the weekly supply chain report. Today their knowledge lives in their head. Genie Agents lets them turn that knowledge (the Genie conversations they have, the questions they know how to ask) into an agent anyone can use. Organizational expertise becomes infrastructure.

Genie App Builder is the self-service analytics dream finally becoming real. Business teams that have been waiting months for engineering to build internal tools can now describe what they want and get a working application, all governed by the same Unity Catalog access controls that apply to everything else on the platform.

Why does it matter for Databricks developers?

Developers can now publish agents as Genie Agents, making their work accessible to a far wider internal audience without building a custom UI. Genie App Builder handles the hosting, scaling, and credential management through Databricks Apps, which means less infrastructure work and more time on the actual logic.

Where to read more

Databricks Blog: Genie One, Genie Agents, and Genie Ontology: databricks.com/blog/introducing-genie-one-genie-ontology-and-genie-agents

App Spaces & Serverless Micro Apps: Governed App Building at Scale

What is it?

Announced at DAIS 2026 alongside Genie App Builder, App Spaces and Serverless Micro Apps are two platform capabilities that make it practical to deploy many apps across an organization without sacrificing governance or incurring prohibitive infrastructure costs.

App Spaces is a new governance boundary inside Databricks Apps. Admins configure resource access, security policies, API scopes, and Unity Catalog permissions once for a group of apps. Every app built inside that Space automatically inherits those guardrails. Governance stops being something you bolt on after the fact and becomes the container apps live in from day one.

Serverless Micro Apps run on a new microVM-based runtime. Each app runs in its own lightweight virtual machine, starts fast when needed, and scales all the way down to zero when idle. The practical outcome: deploying a lightweight departmental tool on Databricks becomes usage-based rather than reserved-capacity-based. Apps that would previously be deprioritized because of always-on cost can now be built and run economically.

Why does it matter for customers?

The combination changes the economics of internal tooling. Previously, a business team wanting a custom app had to justify always-on infrastructure costs. With Serverless Micro Apps that scale to zero, a procurement dashboard or a weekly ops briefing tool costs almost nothing when nobody is using it. App Spaces means the security and governance review happens once at the Space level, not for every individual app, removing a significant bottleneck to scaling self-service app development across departments.

Why does it matter for Databricks developers?

The pattern is: App Spaces establishes the governance boundary, Genie App Builder creates inside it with full awareness of your data and workspace context, and microVM-based Serverless Micro Apps make the whole model economically sustainable at scale. Built apps can now also be listed on Databricks Marketplace and distributed to 20,000+ customers, with IP protected, since customers never see the underlying code.

Where to read more

Databricks Blog: Enabling Governed Vibe Coding for Enterprise Apps: databricks.com/blog/enabling-governed-vibe-coding-enterprise-apps-databricks

Genie ZeroOps: Autopilot for Your Data Infrastructure

What Genie ZeroOps is?

Genie ZeroOps is an autonomous agent for data infrastructure management. It watches your pipelines, detects failures and anomalies, runs experiments in a safe non-production sandbox, and proposes fixes for a human to approve. The goal is a world where the data platform largely runs itself, and humans focus on decisions rather than operations.

Why does it matter for customers?

Data pipeline failures are expensive: missed SLAs, stale reports, cascading downstream issues. The response today usually involves someone getting paged, investigating, and manually patching. ZeroOps compresses that loop significantly by the time a human reviews the proposed fix, the diagnosis is already done and a solution is ready to approve. Teams running lean data operations will feel this most immediately.

Why does it matter for Databricks developers?

For data engineers, ZeroOps is the beginning of a shift from operator to approver. The ambition is that routine pipeline maintenance (schema drift, broken dependencies, performance regressions) gets handled autonomously, while engineers spend their time on architecture, new capabilities, and edge cases the agent cannot handle. It is early (Private Preview), but the direction is significant.

Where to read more

Databricks Blog: Genie One, Genie Agents, and Genie Ontology: databricks.com/blog/introducing-genie-one-genie-ontology-and-genie-agents

Lakebase: The Operational Database Built for the Agent Era

Lakebase was announced as Generally Available at the summit, bringing a fully managed, serverless PostgreSQL database directly into the Databricks Data Intelligence Platform.

The Architectural Problem Lakebase Solves

For decades, enterprise data architectures have maintained a hard separation between operational databases (OLTP), analytical platforms (OLAP), and AI workloads. This separation created data duplication, fragile ETL pipelines, governance gaps, and latency between systems. AI agents in particular suffer, they need low-latency transactional access to update application state, not just analytical read access to historical data. Lakebase eliminates this separation.

Why does it matter for customers?

Until now, organizations building AI applications had to maintain two separate stacks: a transactional database (Postgres, MySQL, Aurora) for the application layer, and an analytical platform (Databricks) for AI and reporting. That meant duplicating data, building ETL pipelines between systems, and managing two sets of governance rules.

Lakebase collapses that into one. Your AI agent can read a customer’s profile, write a decision back to the same database, and have that decision show up in your analytics — all without crossing a system boundary. At 12 million database launches per day, this is clearly landing with real workloads.

Why does it matter for Databricks developers?

Git-style branching alone is a major quality-of-life improvement. Reproducing a production bug locally without touching live data (or testing a schema migration against a real-data clone) is something developers have wanted for years. Database branching, borrowed from tools like Neon, brings that to Databricks natively.

Lakebase Search means you no longer need a separate vector database (Pinecone, Weaviate, pgvector in a separate cluster) for AI retrieval. Hybrid search lives in the same Postgres instance as your operational data, under the same Unity Catalog permissions.

What’s New at DAIS 2026

• Cross-cloud and cross-region disaster recovery
• Git-style database branching — spin up a full-fidelity branch of a live production database in sub-seconds, enabling safe debugging without touching production
• Autonomous database operations — automated maintenance, scaling, and optimization
• Lakebase Search (Beta) — hybrid vector and full-text retrieval built natively into Postgres, with 32× compression enabling billion-plus vector indexes at low cost
• 12 million database launches per day as of the summit announcement

Where to read more

Databricks Blog: Unifying Data and Governance in the Agentic Era: databricks.com/blog/unifying-data-and-governance-agentic-era-whats-new-azure-databricks

Databricks Blog: Announcing Lakebase Launch Partners: databricks.com/blog/announcing-databricks-lakebase-launch-partners

Lakehouse//RT (Beta): Sub-Second Analytics Without a Separate Warehouse

What Lakehouse//RT it?

Lakehouse//RT is a real-time analytics engine delivering sub-100ms query latency at 12,000 queries per second, running directly on Delta and Apache Iceberg tables. No separate real-time database required. The Reyden engine  (built specifically for this) powers the speed improvements.

Why does it matter for customers?

The traditional answer to ‘I need fast analytics’ was to move data into a dedicated real-time warehouse (Druid, Pinot, ClickHouse). That meant another pipeline, another governance layer, another bill, and another system to keep in sync. Lakehouse//RT eliminates that entire category of infrastructure for many workloads.

Healthcare company PointClickCare was cited at the summit: their benchmarks showed Lakehouse//RT running more than a third faster on average than their prior warehouse, with some individual queries running 10× faster, all without standing up a separate system.

Why does it matter for Databricks developers?

If you are building dashboards, operational reports, or AI-driven applications where query latency matters  (think live inventory, real-time customer 360, fraud detection) you now have a path that stays entirely within the lakehouse. No connector to maintain, no separate SLA to manage, no additional security perimeter to govern.

Where to read more

Databricks Blog: Unifying Data and Governance in the Agentic Era: databricks.com/blog/unifying-data-and-governance-agentic-era-whats-new-azure-databricks

LTAP: Rethinking Data Architecture Beyond Traditional ETL

What LTAP is?

LTAP stands for Lake Transactional/Analytical Processing. It is the framework that ties Lakebase and the Lakehouse together into a single governed system where transactional data and analytical data live on one copy of storage, in open formats, with no ETL pipelines between them.

To understand why this matters, a quick history lesson. For roughly 45 years, enterprise data has lived in two separate worlds. OLTP databases  (Postgres, MySQL, Oracle) handle live transactional work: taking orders, processing payments, updating records. OLAP systems (data warehouses, lakehouses) handle analysis dashboards, machine learning, and reporting. The bridge between them is ETL: extract, transform, load. ETL is expensive, slow, error-prone, and means you are always running on two slightly out-of-sync copies of the same data.

LTAP eliminates that bridge. Lakebase stores transactional data directly in Unity Catalog using the same open Delta and Iceberg formats as the Lakehouse. Analytical engines read the same storage layer. Transactional and analytical compute run independently so they do not interfere with each other, but they share one governed copy of data.

“For forty years we’ve lived with a separation between OLTP and OLAP. For the first time, we think we’ve cracked the unification code.”

— Ali Ghodsi, Co-founder and CEO, Databricks (as quoted in Forbes)

Why does it matter for customers?

Previous attempts at merging OLTP and OLAP (usually called HTAP (Hybrid Transactional/Analytical Processing)) were tried by SAP HANA, Oracle, and others. They mostly failed in practice because they forced both workloads into a single engine, trading performance on one side for gains on the other, and because they relied on proprietary storage that locked customers in.

LTAP takes a different approach: unified storage, separate compute. Each workload gets an engine tuned for it. The data is stored once, in open formats. Because it is built on Lakebase (Postgres-compatible) and governed by Unity Catalog, it fits into the toolchains organizations already use.

The practical impact: an AI agent that wants to read a customer’s live order history and reason over six months of purchasing patterns today has to query two different systems and reconcile two potentially out-of-sync copies of the data. With LTAP, there is one copy. The agent reads it, it is current and governed, no reconciliation step.

Why does it matter for Databricks developers?

If you are building AI agents or applications that need to both read live operational data and run analytical queries, LTAP is the architecture you have been waiting for. One platform, one permission model, one copy of data. An agent’s view of a record is the same as the operational system’s view, because they are reading the same file.

Key technical detail: Lakebase now stores data directly in Unity Catalog in Delta/Iceberg format. That means any Databricks engine (Spark, SQL Warehouse, the Reyden engine powering Lakehouse//RT) can query operational data without any transformation step. Agents can write a transaction and have it immediately available for analytics. 

One historical footnote: Databricks acquired Neon (serverless Postgres) and Mooncake Labs in 2025. Those acquisitions are now clearly visible in Lakebase and, by extension, LTAP. The git-style database branching that Lakebase offers (sub-second clones of production databases for safe testing) comes directly from Neon’s approach.

Where to read more

Official LTAP press release: databricks.com/company/newsroom/press-releases/databricks-launches-ltap-first-lake-transactionalanalytical

Databricks Blog: Unifying Data and Governance in the Agentic Era: databricks.com/blog/unifying-data-and-governance-agentic-era-whats-new-azure-databricks

Agent Bricks: The Complete Platform for Building Production AI Agents

What Agent Bricks is?

Agent Bricks is Databricks’ developer platform for building, deploying, and operating AI agents. It was introduced last year, but the numbers from the keynote tell you how it has grown: over 100,000 agents built, over 1 quadrillion tokens processed per year, and production deployments at companies including AstraZeneca, 7-Eleven, Fox Corporation, and Block.

The DAIS 2026 expansion makes Agent Bricks a full-stack agent development environment. That means model choice (all major frontier models, plus a new SpaceX partnership bringing Grok models natively into Databricks), support for every major agent framework (LangGraph, Agno, CrewAI, Claude Code SDK, OpenAI Agent SDKs), secure execution sandboxes, built-in document AI, and a new meta-orchestration layer called Omnigent.

Why does it matter for customers?

Building an AI agent is the easy part. The hard part (the 99% of engineering work that does not show up in demos) is everything else: managing token costs at scale, handling security when agents execute code, evaluating whether agents are actually getting things right, monitoring them in production, and governing what they can access and do.

Agent Bricks now addresses all of that in one place. The alternative is assembling it yourself from a dozen different tools, each with its own governance model and observability story.

Why does it matter for Databricks developers?

A few specifics worth calling out for practitioners:

• Omnigent (Beta): A managed meta-harness for orchestrating agents across different frameworks, models, and tools, useful when you have agents built in different stacks that need to work together.
• Document AI functions: ai_parse_document, ai_extract, and ai_classify are now SQL-callable functions. Databricks claims highest quality and lowest cost on enterprise document benchmarks. If you process PDFs, contracts, or invoices, these are worth benchmarking.
• Secure sandboxes: Isolated VMs for code interpreter tools and subagents, with downscoped Unity Catalog credentials. Safe code execution without a custom security engineering project.
• Custom model training: Using reinforcement learning, Databricks trained a custom data agent competitive with frontier models like Claude Opus on Genie-related tasks, at significantly lower cost per query. Available for customers who want to do the same.

Where to read more

Databricks Blog: Agent Bricks at DAIS 2026: databricks.com/blog/agent-bricks-dais-2026

Unity AI Gateway: Governing AI in Production

What Unity AI Gateway is?

Unity AI Gateway extends Databricks’ existing governance framework to cover the live interactions between AI agents, models, MCP services, and enterprise tools, not just the assets themselves. Think of it as a policy enforcement and observability layer for your entire AI runtime.

At DAIS 2026, four major areas were expanded: 

• cost controls and smart routing (spend caps, workload routing to balance quality and cost), 
• unified runtime governance (enforcing policies at the moment of interaction, not just at design time), 
• production monitoring and investigation (end-to-end trace capture, incident investigation via Lakewatch), 
• and open ecosystem security integrations (including a new partnership with Zscaler).

Why does it matter for customers?

Here is a problem that is starting to hit enterprises running AI at scale: nobody knows what their AI is spending. Models get called by dozens of agents, costs accumulate invisibly, and the first sign of a problem is a surprise invoice. Unity AI Gateway puts hard spend caps and visibility in place before that happens.

The governance angle is equally important. ‘We have a policy that agents cannot access PII without explicit approval’ is meaningless without runtime enforcement. Unity AI Gateway makes it enforced at the infrastructure level, not just documented in a policy doc.

Why does it matter for Databricks developers?

End-to-end trace capture is the feature practitioners will feel most directly. When an agent produces a wrong answer or an unexpected result, you need to replay exactly what happened — what context it had, what it called, what was returned. That kind of forensic capability is what separates production-grade AI systems from demos that occasionally work.

Where to read more

Databricks Blog: AI Governance at DAIS 2026 – Unity AI Gateway: databricks.com/blog/ai-governance-data-ai-summit-2026-whats-new-unity-ai-gateway

Unity Catalog: Business Glossary, Domains, and Metrics

How Unity Catalog Improved?

Unity Catalog received three new semantic layer capabilities that work in concert with Genie Ontology. Business Glossary (Preview) lets teams define authoritative business terms ( ‘ARR,’ ‘active user,’ ‘churned customer’) and connect them to the underlying data assets. Domains (Public Preview) organize the catalog into business-aligned areas so agents and users get scoped, relevant context rather than the entire catalog at once. Metrics turns KPIs into governed, reusable objects that can be called from dashboards, agents, and applications without being redefined each time.

Why does it matter for customers?

These three features solve a problem that sounds boring but is genuinely painful: nobody agrees on what anything means. Finance’s ‘revenue’ and sales’ ‘revenue’ are different numbers. The difference only surfaces in a board meeting. Business Glossary forces that conversation to happen once, records the outcome, and makes it available to every AI system, dashboard, and report automatically.

Why does it matter for Databricks developers?

If you are building agents that need to know what a business term means, Business Glossary is now the authoritative source automatically consulted by Genie Ontology. Metrics as governed objects means you write the definition of ‘monthly active users’ once, certify it, and every agent, notebook, and dashboard calls the same calculation. No more drift between teams.

Where to read more

Databricks Blog: AI Governance at DAIS 2026 Unity AI Gateway: databricks.com/blog/ai-governance-data-ai-summit-2026-whats-new-unity-ai-gateway

OpenSharing: An Open Standard for Sharing AI, Not Just Data

What OpenSharing is?

OpenSharing is the next evolution of Delta Sharing, Databricks’ open data-sharing protocol. Delta Sharing (launched 2021) solved how to share structured data across organizations without copying files, it became the most widely adopted open data-sharing protocol. OpenSharing extends the same principle to the full AI stack: models, agent skills, Genie Agents, and unstructured data can now be shared via a single open protocol.

Announced jointly by Databricks and the Linux Foundation on June 10  (six days before the summit) it is already available on GitHub.

“Delta Sharing proved the industry would choose open over locked-in. OpenSharing extends that principle to the full AI stack, while expanding the cross-platform ecosystem to Iceberg recipients and on-premises providers.”

— Matei Zaharia, Co-founder and CTO, Databricks

Why does it matter for customers?

If your organization collaborates with partners, suppliers, or customers  (and needs to share data or AI capabilities with them) OpenSharing is the plumbing that makes it happen without bespoke integrations. Genie Agent Sharing (Beta) is particularly interesting: you can share a natural-language AI interface with a partner, giving them curated access to your data without exposing proprietary logic, with controls on what they can ask, how much data they can export, and how many requests they can make per day.

Why does it matter for Databricks developers?

Apache Iceberg REST Catalog API support means OpenSharing now reaches consumers who are not running Databricks, no format conversion needed. On-premises storage support (via partners including MinIO, Qumulo, and Everpure) means the protocol reaches data that never moves to the cloud. SecureConnect (Public Preview) removes the networking headache: cross-cloud storage connections without firewall rules or DNS configuration.

Where to read more

Official press release: databricks.com/company/newsroom/press-releases/databricks-announces-opensharing

Databricks Blog: Introducing OpenSharing: databricks.com/blog/introducing-opensharing-next-evolution-delta-sharing-agentic-era

Databricks Blog: New OpenSharing and Marketplace capabilities: databricks.com/blog/announcing-new-opensharing-and-marketplace-capabilities-ai-era

Azure Databricks: New Microsoft Integrations

What is it?

A set of integrations deepening the Databricks-Microsoft partnership, including Genie directly inside Microsoft Teams and M365 Copilot (Beta), an Excel Add-in for self-service analytics (Public Preview), a SharePoint Connector for pulling files into Databricks (Beta), OneLake Catalog Federation (GA) for querying OneLake data without ETL, and CustomerLake, a full Customer Data Platform built natively in the Azure Databricks lakehouse using Profile Agents and Campaign Agents.

Why does it matter for customers?

For organizations already in the Microsoft ecosystem, these integrations mean Genie’s data-grounded intelligence shows up where people already work (in Teams, in Excel) without requiring users to open a new application. CustomerLake is the more strategic announcement: it positions Azure Databricks as a CDP alternative, letting marketing and customer success teams build Customer 360 profiles and run personalized campaigns directly from the lakehouse, without purchasing a separate CDPplatform.

Why does it matter for Databricks developers?

OneLake Catalog Federation GA means you can query Microsoft Fabric data from Azure Databricks without moving it. If your organization uses both platforms, that removes a significant friction point. The Excel Add-in means non-technical users can pull governed lakehouse data into spreadsheets without involving engineering, which should reduce the number of one-off data requests hitting your team.

Where to read more

Databricks Blog: Unifying Data and Governance in the Agentic Era: databricks.com/blog/unifying-data-and-governance-agentic-era-whats-new-azure-databricks

Microsoft Community Hub: Azure Databricks at DAIS 2026: techcommunity.microsoft.com/blog/azure-databricks/azure-databricks-at-databricks-data–ai-summit-2026-updates-and-new-announcements/4528388

Security & Compliance: Identity, Networking, and Certifications

What is it?

A broad security update touching identity management, network architecture, and compliance certifications. Highlights include Automatic Identity Management (AIM) for Microsoft Entra ID now GA on AWS and GCP, AIM for Okta in Public Preview, Context-Based Ingress (CBI) enabling zero-trust access policies, a new Private Network Gateway for secure serverless access to private resources, HITRUST certification across all three major clouds, and FedRAMP High coming to Azure Commercial in summer 2026.

Why does it matter for customers?

AIM removing manual SCIM synchronization is underappreciated. Keeping user groups and permissions in sync between your identity provider and Databricks has historically been error-prone work that either falls through the cracks or consumes significant IT time. AIM automates it with your IdP as the source of truth. CBI is the enabler for safely deploying Genie to broader audiences (including external users) without opening the entire workspace to the public internet.

Why does it matter for Databricks developers?

FedRAMP High and the AWS GovCloud expansion for AI products (Apps, Model Serving, Genie, Genie Code) open Databricks’ AI capabilities to U.S. federal agencies and their partners for the first time in earnest. If you build on Databricks for government customers, this is a significant door opening.

Where to read more

Databricks Blog: Platform Security and Compliance at DAIS 2026: databricks.com/blog/whats-new-databricks-platform-security-and-compliance-data-ai-summit-2026

Databricks Free Edition: Five More Products, No Cost

What is it?

Databricks Free Edition (now used by more than 500,000 people) expanded to include five previously paid products at no cost: Genie Code, Serverless GPUs, Lakebase, Agent Bricks, and Lakeflow Designer.

Why does it matter for customers?

The full toolkit for building data pipelines, AI agents, and agentic applications is now accessible without a paid contract. For teams evaluating Databricks, or for practitioners learning on their own, the barrier to getting hands-on with production-grade tooling just dropped substantially.

Why does it matter for Databricks developers?

Serverless GPUs in the free tier means you can train and fine-tune models without needing cloud credits. Lakebase and Agent Bricks free means you can prototype an entire agentic application (operational database, agent logic, analytics) without spending anything. It is a meaningful ecosystem investment.

Quick Reference: All Major Announcements at a Glance

The 5 Things Every Enterprise Should Take From DAIS 2026

1. Semantic modeling just became urgent

Genie Ontology and Business Glossary are powerful, but only as good as the business definitions you feed them. If your organization has never formally defined what ‘active user’ or ‘ARR’ means, that work is now a prerequisite for competitive AI, not a nice-to-have.

2. Governance is moving from the catalog to the runtime

Unity AI Gateway marks a shift: it is no longer enough to govern what AI assets exist and who can access them. Production AI requires enforcement at the moment of interaction. If your governance strategy stops at the catalog, it is incomplete.

3. The OLTP / OLAP split is ending

Lakebase combined with Lakehouse//RT means you can run transactional writes and sub-second analytical reads from the same governed lakehouse. For organizations building AI applications, this eliminates an entire tier of data infrastructure and the fragmentation that comes with it.

4. The shift from ‘building AI’ to ‘operating AI’ is real

Every major enterprise capability at DAIS 2026 (ZeroOps, Unity AI Gateway monitoring, Agent Bricks evaluation and sandboxing) addresses the operational challenge, not the build challenge. If your AI strategy is still primarily focused on which model to pick, you are solving for the wrong bottleneck.

5. Open standards remain Databricks’ long-term bet

Delta, MLflow, Delta Sharing, and now OpenSharing are all open-sourced and foundation-hosted. For enterprise organizations, this matters practically: it limits lock-in and creates a larger partner ecosystem. The Linux Foundation hosting of OpenSharing reinforces that this is infrastructure, not a product feature.

About Datapao

Datapao is a Databricks-focused data and AI consulting and training company operating in the UK and across Europe, with delivery hubs in the UK and Central Europe. The company helps organisations design, build, and operate modern data platforms while enabling internal teams through hands-on Databricks training. Our approach combines deep engineering expertise with structured knowledge transfer to ensure sustainable, long-term adoption of data and AI.

Get in touch to discuss where to start.