Most of a company’s data was never built to be searched. Generate reaches it anyway — without moving it, and without a per-token bill that grows every quarter.
A joint educational white paper by NetApp® and Iterate.ai.
Somewhere in every large company sits a warehouse of contracts, claims files, service tickets, and inspection reports that nobody has opened in years. It isn’t lost. It’s filed and stored away on a share drive, in a database, inside a document repository behind the right permissions. Filed is not the same as usable. If no one can ask it a question, it might as well not exist.
That gap between “stored” and “searchable” is dark storage: 80–90% of enterprise data by most estimates. It sits there while teams rebuild answers by hand, and while a separate problem compounds it — the cost of running AI against the data a company can reach.
Public AI tools solve the second problem by making a model easy to reach, but don't solve the first. They still can’t see a company’s private repositories, and every query against them adds to a token bill that scales with use rather than with value delivered. A model that can’t reach the filing cabinet, billed by the question, resolves nothing about dark storage. It just makes asking questions from the outside more expensive.
The two problems share a root cause: the AI runs somewhere other than where the data lives. Every workaround — copying files to a staging environment, building a custom pipeline per data source, negotiating a bigger token allowance — treats the symptom. A platform built to sit inside the company’s own infrastructure, next to the data that’s already there, removes the reason those workarounds exist.
That’s what Generate is built to do — the subject of the rest of this brief.
Generate is a private AI platform that runs fully inside a company’s own infrastructure — the models, the retrieval layer, and the vector database all live where the data already lives, instead of the data traveling out to a public model. Nothing about the security perimeter changes, and nothing leaves the environment to get an answer.
Generate’s connectors read documents, databases, and applications where they already sit. Paired with ONTAP, the data a company has been storing for years becomes the same data its agents can query today — no migration, no separate copy, no new place for it to live.
A fixed-cost, contained platform changes who gets to run a pilot. A team doesn’t need a budget line that grows with every query, and it doesn’t need months of integration work before the first agent sees real data. It needs one connector pointed at the repository that’s been sitting dark, and a place to run the model that doesn’t require moving anything out the door.
Where this fits in the series. This brief describes the Generate platform as the private-AI layer that sits on NetApp’s infrastructure. It does not restate the AIPod Mini’s hardware specifications or pricing, which are covered in their own briefs.
The End of “Per Token” Costs — how token-based pricing compounds against agentic workloads, and what a fixed-cost alternative looks like.
What’s in Storage — a closer look at what dark, unstructured enterprise data actually contains.
Read the full series at iterate.ai/partners/netapp/private-ai-education-and-certification-program.
A joint educational series on private AI and the AIPod Mini. NetApp — the governed data-control layer. ONTAP stores, secures, and keeps your dark storage private and on-prem. Iterate.ai — the private intelligence layer. Generate turns that data into agents, answers, and decisions in minutes.
