Written in partnership with NetApp
Most AI projects don’t stall because of the model. They stall because the data underneath it isn’t ready. Here’s the four-stage journey from scattered data estate to AI-ready enterprise and, why Proact and NetApp are the fastest path to get there.
The short version
Enterprise AI projects fail on data, not models. Before any organisation can reliably use AI, their data estate must be discovered, classified, governed, and connected to an LLM. NetApp BlueXP and AIDE handle the discovery and enrichment layer. Microsoft Purview governs M365 data. BlakYaks, a Proact company, builds and operates the Azure cloud native platform the AI workloads run on. Proact IT UK designs, deploys and manages the full stack as one system, typically delivering a working AI query capability within six weeks, with zero data migration required.
Why do enterprise AI projects fail?
Enterprise AI projects most commonly fail because the underlying data estate is not ready, not because the AI technology itself is immature.
Every organisation is under pressure to deploy AI. Boards are asking for it. Vendors are selling it. But most IT and data teams know the reality: the data their organisation holds is scattered, inconsistently structured, and nowhere near ready to be queried by an AI system.
The root cause is not the model, the compute, or the budget. The root cause is that AI cannot reliably query data it cannot find, classify, or access safely and most enterprise data estates have never been built with AI in mind.
AI is not a model problem. It’s a data problem. Your data estate is already the answer. It just needs the right foundation.
What does an unready data estate look like?
An enterprise data estate that is not AI-ready typically has four characteristics:
- Data is distributed across on-premises storage arrays, Microsoft 365, SaaS platforms, and cloud environments with no unified view.
- Data is unclassified. Sensitive data, operational data, and stale data coexist with no consistent tagging or categorisation.
- Governance is absent. There is no policy layer controlling which users or systems can access which data.
- There is no connective tissue. Storage, governance, and AI applications are managed separately, by different teams, with no one responsible for making the whole system work.
Each of these problems is solvable. But they need to be solved in the right sequence, with the right technology, and with someone responsible for making all the pieces work together.
What is the four-stage journey to AI readiness?
The journey from a scattered enterprise data estate to a working AI capability runs through four sequential stages: discover, enrich, govern and connect.
Each stage builds on the previous one. Skipping a stage, which many organisations attempt, is the most common reason AI projects produce unreliable outputs or fail to reach production.
Stage 01: Foundation
How do you discover what data you have?
NetApp BlueXP Classification scans and catalogues the ONTAP data estate across on-premises and cloud environments. BlueXP applies classification policies automatically without requiring manual tagging or data migration. Organisations running NetApp ONTAP gain full estate visibility, typically within days of deployment.
Stage 02: Intelligence
How does NetApp AIDE make data AI-ready?
NetApp AIDE (AI Data Engine) builds a global, enriched metadata catalogue across the entire data estate. AIDE tags, structures, and contextualises data so it is ready to be connected to a large language model without moving data from its existing location. AIDE integrates directly with NVIDIA NeMo Retriever, enabling natural language queries against enterprise data in production.
Stage 03: Governance
Why does data governance come before AI connection?
Connecting AI to an ungoverned data estate creates significant compliance and security risk. Microsoft Purview applies sensitivity labels, data loss prevention policies, and access controls specifically across the Microsoft 365 estate, ensuring AI only queries M365 data it is authorised to access. BlakYaks and specialist Microsoft Azure partner, builds the Azure cloud native platform that the AI workloads run on, using DevSecOps practices, Infrastructure as Code, and pre-engineered Azure Accelerator Packs including an AI Foundations pack.
Stage 04: Outcome
What does a working enterprise AI query look like?
With data discovered, enriched, and governed, connecting AI is straightforward. NVIDIA NeMo Retriever integration enables users to prompt the enterprise data estate in natural language, live, accurate, and safe. Microsoft 365 Copilot, custom LLMs, and AI-assisted search all operate reliably on a properly prepared data foundation. This is the outcome most organisations have been promised but few have delivered.
For organisations moving beyond a proof of concept into production inferencing, the infrastructure you run it on becomes the next constraint. NetApp AI Pod is the validated stack purpose-built for this, pairing ONTAP AFF storage with NVIDIA GPU compute in a single architecture sized on real workload evidence. AI Pod Mini provides a right-sized entry point for departmental use cases. Proact sizes both using data collected during the DII assessment, so the hardware conversation is grounded in evidence, not estimates.
Why is NetApp the fastest path to AI on your data?
NetApp provides the fastest route to an AI-queryable data estate because NetApp ONTAP is already deployed in most enterprise environments, making enrichment a configuration exercise, not a migration project.
NetApp’s approach is built on a clear strategic premise: organisations should not have to rebuild their data architecture to deploy AI. The ONTAP footprint that already exists in an enterprise environment becomes AI-ready through BlueXP Classification and AIDE, without ripping out existing infrastructure or moving data.
What are the key NetApp capabilities for enterprise AI?
| Capability | What it does | Status |
|---|---|---|
| BlueXP Classification | Scans and classifies the ONTAP estate across on-prem and cloud environments. | Live today |
| AIDE (AI Data Engine) | Builds a global enriched metadata catalogue ready for LLM connection. | Live today |
| NVIDIA NeMo Retriever | Mature RAG integration enabling natural language queries against enterprise data. | Production-ready |
| ONTAP Foundation | The existing storage infrastructure that BlueXP and AIDE operate on. | Already deployed |
| Data Infrastructure Insights | Collects 8,640+ real workload data points over 30 days. Vendor-agnostic, free for 90 days. Sizes the AI Pod conversation on evidence. | Live today |
| NetApp AI Pod/AI Pod Mini | Validated AI compute stack pairing ONTAP AFF storage with NVIDIA GPU infrastructure. Sized on DII findings. | Production-ready |
What does NetApp not cover?
NetApp BlueXP and AIDE are strongest on the ONTAP footprint. Microsoft 365 data and SaaS data sources sit outside NetApp’s direct scope and require a separate governance layer. This is not a limitation of NetApp’s approach. It is simply the reality of where enterprise data lives today, spread across multiple environments that no single vendor fully owns.
What is the role of Proact IT UK in an enterprise AI data project?
Proact IT UK acts as the orchestration layer across the full enterprise AI data stack, designing, deploying, integrating, and managing the system that individual vendors cannot build alone.
No single vendor solves the complete enterprise AI data challenge. NetApp BlueXP and AIDE cover the ONTAP data estate. Microsoft Purview governs the M365 estate. NVIDIA NeMo Retriever provides the AI query layer. BlakYaks, builds and operates the Azure cloud native platform that the AI workloads run on, covering DevSecOps, Infrastructure as Code, and enterprise Azure landing zones. Proact IT UK is responsible for making all of these components work as one coherent system.
What does Proact IT UK do in practice?
- Design the architecture: Proact IT UK maps the current data estate, identifies gaps, and designs an architecture connecting NetApp, Microsoft, and the AI layer before any deployment begins.
- Deploy the stack: NetApp BlueXP and AIDE for ONTAP data classification, Microsoft Purview for M365 access controls, BlakYaks Azure cloud native platform and AI Foundations accelerator pack, NVIDIA NeMo Retriever integration, and NetApp AI Pod or AI Pod Mini sized on DII workload evidence for production inferencing, all deployed, tested and validated by Proact IT UK.
- Cover the gaps: Proact IT UK bridges Microsoft 365 and SaaS data into the NetApp intelligence layer so the whole estate is covered, not just the storage.
- Manage it ongoing: Proact Managed Services keeps the stack running, monitored and current as the AI landscape evolves.
Proact IT UK turns fragmented enterprise data into something AI can safely and reliably use, no matter where it lives.
How Quickly Can Proact IT UK Deliver a Working AI Capability?
How quickly can Proact IT UK deliver a working AI capability?
Organisations running NetApp ONTAP can typically reach a working AI query capability within six weeks. The Proact IT UK programme requires no data migration and no architectural rebuild. The four delivery commitments are:
- Weeks to first AI query – not months to a proof of concept.
- Zero data movement – data stays where it already lives in the ONTAP estate.
- No architecture change – BlueXP and AIDE enrich the existing infrastructure.
- Managed ongoing – Proact IT UK operates the stack as a managed service from day one.
How do you know if your organisation is ready to start?
Two diagnostic questions tell you where your organisation sits on the AI readiness journey and what the right starting point is.
Most organisations do not need to have solved AI before starting the conversation. They need honest answers to two questions that surface where the real work is.
Where does your data live today?
If the majority of enterprise data lives in NetApp ONTAP, the organisation is closer to AI readiness than most. BlueXP Classification and AIDE can begin enriching that estate immediately. If data is spread across multiple environments with no consistent structure, the starting point is a data readiness assessment, which Proact IT UK can scope and deliver before any vendor deployment begins.
Who owns the AI agenda in your organisation?
If the CTO, CISO, or a named data owner is driving the AI programme internally, there is a real sponsor to work with. If nobody owns the AI agenda, the first deliverable is a governance and sponsorship framework, establishing who is responsible for data, who owns the AI outcomes, and what success looks like before any technology is deployed.
Frequently Asked Questions
Why do enterprise AI projects fail?
Enterprise AI projects most commonly fail because the underlying data estate is not ready. Data is scattered across on-premises storage, Microsoft 365, SaaS platforms, and cloud environments with no consistent classification, tagging, or governance. AI models cannot reliably query data that has not been found, enriched, and governed first. The model itself is rarely the problem.
What is NetApp AIDE and what does it do?
NetApp AIDE (AI Data Engine) builds a global, enriched metadata catalogue across an organisation’s entire data estate. AIDE tags, structures, and contextualises data so it is ready to be connected to a large language model without moving data from its existing location. AIDE integrates directly with NVIDIA NeMo Retriever to enable natural language queries against enterprise data in production.
How does Microsoft Purview fit into an enterprise AI architecture?
Microsoft Purview provides the data governance layer for the Microsoft 365 estate. Purview applies sensitivity labels, data loss prevention policies, and access controls so that AI systems only query M365 data they are authorised to access. Without a governance layer like Microsoft Purview, connecting AI to an M365 environment creates significant compliance and security risk.
How long does it take to make an enterprise data estate AI-ready with NetApp?
Organisations running NetApp ONTAP can typically reach a working AI query capability within six weeks. NetApp BlueXP Classification and AIDE operate on the existing ONTAP infrastructure without requiring data migration or architectural rebuilds. Proact IT UK designs, deploys, and integrates the full stack as a single managed programme.
What is the role of Proact IT UK in an enterprise AI data project?
Proact IT UK acts as the orchestration layer across the full enterprise AI data stack. Proact IT UK designs the architecture, deploys NetApp BlueXP and AIDE for ONTAP data classification, connects Microsoft Purview for M365 governance, integrates NVIDIA NeMo Retriever and manages the system on an ongoing basis. BlakYaks is a specialist Microsoft Azure partner that builds and operates the Azure cloud native platform the AI workloads run on, using DevSecOps, Infrastructure as Code, and pre-engineered Azure Accelerator Packs including an AI Foundations pack.
Do I need new hardware to run enterprise AI in production?
For early-stage querying and proof of concept work, existing infrastructure is often sufficient. Moving to production inferencing at scale is a different question. NetApp AI Pod is the validated compute stack designed for this, combining ONTAP AFF storage with NVIDIA GPU infrastructure in a single, tested architecture. AI Pod Mini provides a right-sized entry point for departmental use cases. Proact uses the Data Infrastructure Insights assessment to size the right option based on your actual workloads, so there are no surprises when the hardware conversation comes.
Written by…
Andy Smith
Chief Technology Officer
Andy Smith is CTO at Proact IT UK, a specialist IT managed services provider in enterprise storage, hybrid cloud, and data infrastructure, since 1994. Andy leads Proact IT UK’s commercial and technology strategy across Storage, Data & AI, Hybrid Cloud, Operational Resilience, Modern Work and Managed Services. Proact IT UK is a NetApp partner and Microsoft solutions partner operating across enterprise and public sector organisations in the UK.