By Rad Anandakumar, Practice Lead, Data and AI, Avocado Consulting
In short: AI assistants built into individual applications can only see their own system's data, so they can't answer the cross-system questions that matter most. A governed data lakehouse gives AI a single, trusted and secure foundation to reason over, which is why it belongs at the centre of any enterprise AI strategy.
Every software vendor is now an AI vendor. Salesforce has Einstein, Microsoft has Copilot, and your ERP, service desk and HR platform have all added an assistant to their interface. The pitch is simple: switch it on, and intelligence appears.
A growing number of practitioners, ourselves included, are placing a different bet. AI embedded in individual applications won't, on its own, deliver the enterprise-wide intelligence organisations need. There has never been a free lunch in data, and AI hasn't changed that. A data lakehouse remains the critical layer to point your AI at.
Why isn't application AI enough?
Application AI can only reason over the data in its own system. An AI built into Salesforce can only see what Salesforce sees. The assistant in your ERP knows your invoices, but not the support tickets that explain them. The one in your service desk knows your complaints, but not the contracts that govern them. Each is intelligent about its own silo and blind to everything else.
Yet the questions that matter to a business rarely sit inside a single system:
- Which customers are at risk, and why?
- Where is margin leaking across the order-to-cash cycle?
- Which contract obligations are we failing to meet?
Answering these needs context that no single application holds. The result is what we call point AI: many narrow assistants, each clever in isolation, none of which can see the enterprise. Stitching them together after the fact doesn't produce enterprise intelligence. It produces a committee of strangers.
What is a data lakehouse?
A data lakehouse is a data architecture that combines the low-cost, flexible storage of a data lake with the governance, quality and structure of a data warehouse. It lets an organisation store, refine and trust all of its data in one place.
The lakehouse solves an old problem. It's the latest chapter in a 30-year effort to give organisations a single, trustworthy view of themselves.
Data warehouses gave us structure and governance. But they were rigid and expensive, and they struggled with the unstructured data that now dominates the enterprise: documents, images, audio and telemetry.
Data lakes answered with cheap, limitless storage for every kind of data. Without governance and quality controls, though, most became data swamps. They were technically full of data, but vast, ungoverned and too untrusted to use.
The lakehouse resolved the trade-off. It combines the openness and scale of the lake with the governance, quality and structure of the warehouse. Its layered (medallion) architecture refines raw data in steps, from Bronze (raw) to Silver (transformed and conformed) to Gold (decision-grade), under unified governance and security. For the first time, a single foundation can hold all of an organisation's data and still be trusted.
AI is simply the next workload this foundation was built to serve. The same architecture now carries one more refinement: an AI-optimised Gold layer of governed, contextualised data shaped specifically for large language models, sitting between the lakehouse and the AI that consumes it.
Why does AI need a data lakehouse?
Because AI is only as accurate, secure and useful as the data beneath it. A lakehouse gives AI three things application assistants can't provide on their own: quality data, access control and business context.
AI is only as good as the data it's given
Garbage in, garbage out still applies. AI should reason over relevant, governed, high-quality data, not whatever happens to live in one system. The medallion architecture exists to turn raw, messy inputs into decision-grade information. It's why so many organisations are disappointed by their AI's accuracy: they've pointed sophisticated models at ungoverned data.
Different users must see different things
A shared assistant that exposes everything to everyone is a privacy incident waiting to happen, which is exactly why so many organisations quietly switch their AI off. The lakehouse enforces access at the data layer, so AI can give each user an intelligent view of only the documents and data they're entitled to see.
Context is everything
AI needs data in the right format, with the categorisations, classifications and business context that make it meaningful. Without that semantic layer, you get plausible-sounding point answers. With it, you get enterprise intelligence. Context is the difference between an AI that sounds informed and one that is.
The bottom line
Application assistants will keep improving, and they have their place. But the organisations that win with AI will be the ones that resist the free lunch and do the foundational work: a governed, contextualised and secure lakehouse that all of their intelligence can point to. The lakehouse isn't a precursor to your AI strategy. It's the foundation of it.
Frequently asked questions
What is the medallion architecture?
It's a layered way of refining data inside a lakehouse. Raw data lands in Bronze, is cleaned and conformed in Silver, and becomes decision-grade in Gold. Many organisations now add an AI-optimised Gold layer shaped for large language models.
What's the difference between a data lake and a data lakehouse?
A data lake stores any kind of data cheaply but has little governance, so quality and trust suffer. A data lakehouse adds warehouse-style governance, structure and security on top of lake storage.
Do I need a lakehouse before using Copilot or other application AI?
No. Application assistants are useful within their own systems. But to answer questions that span several systems, or to control who sees what across them, you need a governed foundation such as a lakehouse.
Can a lakehouse control what AI shows each user?
Yes. Access is enforced at the data layer, so AI only surfaces the documents and data each user is entitled to see.
Ready to build the layer your AI needs?
Avocado's AI and Data practice helps organisations turn fragmented systems and ungoverned data into a single, trusted lakehouse foundation, plus the AI-optimised layer that sits on top of it. If your AI is underperforming, exposing data it shouldn't, or trapped inside one application, the problem is almost certainly the foundation beneath it.
Book a discovery session to talk about building your enterprise AI foundation, explore our Data Lakehouse service.
About the author
Rad Anandakumar is Practice Lead, Data and AI at Avocado Consulting. He founded AI Consulting Group, which joined Avocado in 2026.
