Autonomous AI agents require secure, high-fidelity enterprise data to perform complex tasks that often require multi-step workflows. But, the most valuable enterprise information often remains trapped across disconnected, multi-modal silos. Lacking a unified platform to synthesize this fragmented data, organizations are forced into piecemeal solutions that trigger severe security risks, soaring cloud and AI token expenses, and frequent model hallucinations due to fragmented data that lacks semantics or context.
VMware Tanzu is introducing AI-ready data foundations to solve these challenges and pave the way for agentic AI success. A native capability of Tanzu Platform, an AI-ready data foundation operates as a centralized data-to-AI lifecycle control plane and environment, enabling you to unify mixed data formats, add a comprehensive semantic layer, and enforce strict governance to deliver secure, context-rich data products directly to AI-agents. Armed with AI-ready data foundations, your IT teams no longer have to stitch together complex, standalone tools. With a pre-engineered, unified environment for data ingestion, transformation, governance, and access, Tanzu enables you to provide agents with secure, cost-efficient access to AI-ready data. AI can benefit from AI-ready data foundations in Tanzu Platform help enterprises adopting agentic AI:
- Break down silos without having to move data with multi-modal ingestion and federated querying
- Power AI agents and spend tokens more efficiently with context-rich data, delivered through high-precision parsing
- Ensure agent access to trusted data with curated data products and a ticketless marketplace
- Achieve airtight security and governance, and comply with local data sovereignty regulations with Tanzu Platform for agents and AI-ready data deployed to your private cloud environment
A Pre-engineered Platform for AI Agents with Low TCO
Setting up an AI-ready data platform via a “DIY” approach is extremely difficult, requiring organizations to stitch together fragmented standalone utilities for ingestion, vector search, analytics, and governance. This tool sprawl drives up software licensing fees and creates massive ongoing engineering maintenance overhead. Furthermore, relying on public cloud data platforms as your primary environment for agentic AI workloads can be cost-prohibitive due to AI compute premiums, frequently up to 2x to 5x more expensive than typical IaaS compute costs, as well as the unplanned data egress fees, duplicate storage footprints, and increased network demands caused by multi-step agentic workflows.
AI-ready data foundations in Tanzu Platform deliver a highly predictable cost structure with a lower TCO (total cost of ownership) thanks to streamlined architecture and operations, such as:
- Zero Tool Sprawl: By automatically incorporating ingestion, transformation, cataloging, and query engines, along with an integrated marketplace for publishing data products, all within a single pre-engineered workflow in Tanzu Platform, AI-ready data foundations reduce weeks-to-months of complex manual integration down to a guided setup that takes a few hours. This dramatically lowers software licensing and ongoing maintenance overhead.
- Lower TCO Than Public Cloud: Leveraging Tanzu Platform on VMware Cloud Foundation (VCF), AI-ready data foundations deliver a fully on-prem solution purpose built for delivering data for AI workloads, eliminating the runaway public cloud billing spikes, duplicate storage fees, and unpredictable data egress charges typical of public cloud-based data platforms.
- One Coherent Stack: Deployed directly alongside Tanzu Platform agent foundations, where developers can quickly build and run secure AI agents, AI-ready data foundations serve as the native data layer. Tanzu Platform gives organizations a single, unified stack to develop and run agents and feed them governed, AI-ready data, completely eliminating the need to stitch separate data vendors into their AI infrastructure.
Transform Raw Data Into Intelligent Data Products For AI Agents
Delivering trusted data to AI-agents presents massive operational and architectural challenges. Valuable enterprise data is typically locked away in disconnected silos, spanning structured legacy databases, semi-structured JSON files, and a massive amount of unstructured data (e.g. PDFs), which accounts for 80–90% of all data.
Traditional data pipelines are ill-equipped to synthesize this fragmented, multi-modal data landscape. Without a unified platform or semantic layer, enterprises are forced to stitch together point solutions (for ingestion, semantics, governance, publishing, etc.) and feed raw data directly into LLMs to make sense of it all. This can lead to astronomical token consumption, for both data ingestion and queries, and frequent model hallucinations.
Here’s an in-depth look at the core features of AI-ready Data foundation and the benefits they deliver to your AI projects.
Break down silos without having to move data with multi-modal ingestion and federated querying
Organizations have traditionally been forced to undergo lengthy, costly, and highly risky data migration projects to centralize their data, with teams often copying structured databases, JSON logs, and unstructured PDFs into a central public cloud data lakehouse just to extract value. Duplicate storage fees, high premiums for AI compute resources, unplanned data egress fees, and network costs make public cloud platforms cost-prohibitive for multi-step agentic workflows. AI-ready data foundations unify access across all data silos while optimizing TCO:
- Multi-Modal Data Ingestion: Through a single platform, the system connects directly to structured sources (such as S3 and GCS object storage, as well as Postgres and Oracle via JDBC), semi-structured data (JSON, CSV, XML, Excel), and unstructured PDFs.
- Federated Querying: Instead of moving physical data, the query engine is brought directly to where the data lives. AI agents can query across all data types simultaneously in a single, unified workflow, eliminating double-storage fees, bypassing regulatory risks, and providing a holistic, real-time view of the enterprise.
Power AI agents and spend tokens more efficiently with context-rich data, delivered through high-precision parsing
When organizations attempt to feed unstructured data into AI applications, they typically rely on standard parsing tools that perform simple, text extraction. This traditional “simple approach” is inherently flawed and not AI-ready because it flattens tables, strips away visual layouts, and loses document hierarchy. Without such structural metadata, agents can’t understand the relational meaning of the data (i.e. figuring out whether a specific value belongs to a column header or grasping that a line of text is a main title, and not a footnote). To make up for this, organizations are forced to feed massive amounts of unrefined, raw datasets directly into LLMs to re-derive meaning, which inflates “token-burn” and leads to astronomical query costs.
An AI-ready data foundation implements an intelligent, structured approach to pre-process, curate, and index multi-modal data, adding structure to unstructured data before it ever reaches an AI agent:
- High-Precision Parsing and Scrubbing: Built-in tools extract structured, business-relevant details, such as text, tables, layout, and hierarchy, from unstructured documents (e.g. native or scanned PDFs). Primary cleansing and sanitation of data (e.g. remove bullets, erratic spacing, etc.) is done automatically without requiring custom engineering.
- Context Preservation: The platform attaches structural metadata and relational context to each visual element. The structural metadata then is bound directly to text-to-vector embeddings during indexing, preserving the exact layout structure and ensuring that the AI agent understands relationally how data elements relate to one another (such as knowing that a specific table cell is bound to its parent title).
- Token Usage Optimization: An AI-ready data foundation does the heavy lifting upstream. Not only does it negate the need to expand LLM token costs during data ingestion, by pre-processing and indexing the data with rich metadata upstream, Tanzu Platform makes sure that AI agents don’t just read massive blocks of raw text; they understand its true meaning, drastically reducing hallucinations, eliminating incorrect outputs, and using tokens more efficiently during active query cycles.
Ensure agent access to trusted data with curated data products and a central marketplace
To be successful, it is imperative that AI initiatives are fueled only by trusted data, which is easier said than done. Operational silos often make it difficult for organizations to have a single, trusted version of truth. In addition, connecting AI-agents and applications to legacy enterprise databases or curated datasets can be very complex, often requiring significant custom coding such as manually building and maintaining bespoke REST APIs, writing custom vector pipelines, or hand-crafting tool-calling wrappers. AI-ready data foundations in Tanzu Platform greatly simplify this process with its Data Product model and Tanzu Platform’s self-service marketplace:
- Curated Data Products: Every ingested source carries inherent structure, tags, indexing, and business context. Data engineers can easily build, clean, mask, sanitize, and enrich these sources using a built-in SQL Editor that offers an advanced, notebook-like flow to build “data products.” They can also craft new data products by joining or transforming other data products for specific enterprise use cases. This guarantees that AI-agents, as well as any other data consumers, consume the exact same verified, context-rich information, eliminating the risk of agents acting on corrupt, outdated, or unverified data.
- Embedded Model Context Protocol (MCP) Server: AI-ready data foundations are embedded with an MCP server, allowing for the binding of governed data products to AI agents. Once a data product is published, any MCP-compliant AI agent or application can then dynamically discover and query data out of the box and can bind to database connections instantly and securely, completely bypassing the need to maintain bespoke wrappers or custom APIs.
- Centralized Data Marketplace: To democratize data access and eliminate the waiting period for manual granting of data access, AI-ready data foundations leverage Tanzu Platform’s unified, self-service marketplace which acts as a secure, friction-free location for trusted and versioned data products. The marketplace provides a machine-readable catalog (searchable by natural language by AI-agents as well as human data consumers), where autonomous agents can discover, understand, and securely access verified data products with intact context. This transitions data access from a ticket-driven manual process to an automated, self-service flow controlled upstream by data preparers.
Achieve airtight security and governance, and comply with data sovereignty regulations with a private cloud environment
Deploying autonomous AI agents on public cloud services introduces severe security and compliance risks, especially for highly regulated industries like banking, healthcare, and government. Sending sensitive, proprietary data or Personally Identifiable Information (PII) to public cloud LLMs can lead to potential IP leakage and violation of strict regional data privacy mandates (such as the EU AI Act) and local data sovereignty laws. Furthermore, establishing a uniform way of governing access and defining “agent identity” is incredibly complex, creating a risk that agents might expose sensitive information to unauthorized users. An AI-ready data foundation addresses these concerns by shifting security controls upstream:
- Upstream Access Controls: Data and platform engineers manage the marketplace, applying strict, role-based access controls scoped to specific Tanzu Platform organizations, spaces, and groups. Access rights are verified before the data ever reaches the AI agent or model, transforming security from a reactive to a proactive process.
- End-to-End Lineage: Every transformation, data ingestion, and publication is recorded. The system maintains a complete audit trail of where data was loaded from, its raw source, and what changes occurred, satisfying compliance teams with a verifiable audit trail.
- Sovereign Private-Cloud Deployment: The entire data, compute, and control environment can run on VCF private-cloud infrastructure. Data ingestion, parsing to add context, and all the necessary compute can all take place in VCF, and the resulting metadata, along with published data products, never leaves your secure physical boundaries, providing absolute physical, operational, and jurisdictional control.
Ensure that your AI investments are providing tangible returns with AI-ready data foundations from VMware Tanzu
AI-ready data foundations in Tanzu Platform transform data from a fragmented, hard-to-access IT asset into a readily available, securely governed corporate data products that fuel today’s agentic AI initiatives. Running securely within a sovereign private cloud, AI-ready data foundations in Tanzu Platform automate unstructured data parsing, maintain a robust semantic layer, standardize agent connections via an embedded MCP server, and provide a centralized Data Marketplace with strict access controls.
Stop wasting time moving and trying to understand data and start acting on it with absolute confidence with Tanzu.
Learn more about Tanzu Platform, the agent platform for VMware Private AI Cloud or contact the Tanzu sales team
