Do I need Snowflake and Cortex, or just this one export?

What is Snowflake built for?

Snowflake is a cloud data platform: you load your organisation's data into it, and it stores and queries that data at a scale and reliability that a spreadsheet cannot approach. Permissions, governance, sharing between teams, and consistent performance across very large tables are the problems it solves, and it solves them well. For a company that needs one trustworthy source of truth across many systems, it is a serious and reasonable choice.

Everything below is about a much smaller problem.

What does Cortex add?

Cortex Analyst is Snowflake's natural-language layer, and it is the part that sounds closest to what Paperswift does. You ask a question in plain English, and it produces SQL against your warehouse rather than making you write it. Snowflake Intelligence extends the same idea into a no-code interface that can work across documents as well as tables.

For a large organisation this is valuable: it takes people who understand the business but not SQL and lets them query governed company data directly, without joining an analyst's queue.

What has to be true before you can ask anything?

The plain-English part is the last step, not the first.

Your data has to be in Snowflake. That means an ingestion path — loading, and usually keeping it loaded and fresh.

Then a semantic model has to exist. Cortex Analyst relies on a file describing what your tables and columns mean in business terms, and the accuracy of its answers tracks the quality of that description. Someone has to write it, and keep writing it as the data changes. Published accounts of production deployments describe accuracy improving substantially as the semantic model, verified queries, and access controls are refined over time. The good results are earned, not switched on.

None of that is wasted effort at organisational scale. It is a great deal of effort for one CSV.

What does each one cost to set up and run?

Snowflake bills compute by consumption, in credits, and this is where costs surprise people.

Each step up in warehouse size consumes credits substantially faster than the one below. An inefficient query can run far longer than it needs to, multiplying what it consumes. Several dashboards can each quietly run their own full scans of the same tables. And some serverless features draw credits outside the warehouse view, so they do not appear where an administrator would naturally look. Snowflake provides resource monitors to cap and alert on this, and the existence of that tooling is the clearest evidence that spend can otherwise get ahead of expectations.

The pattern is the same one the AI assistants have, arriving through a different door: the person asking the question is rarely the person who can see what it costs, and the feedback comes later, on a bill rather than in the moment. Someone asking a business question through a no-code interface has no way to know that the phrasing they chose scanned a very large table.

Paperswift has no meter at all. Questions are unlimited on every tier including the free one, the query runs on your own machine, and the model only ever receives your column names and their types — so a small file and a large one cost the same to ask about, which is nothing. The free tier shows every chart on screen; taking results back out (downloading, saving a chart, exporting) is what the paid tier adds. Current figures are on Snowflake's pricing page and ours.

What actually leaves your computer?

With Snowflake, the data lives in the platform by definition; that is the point of a warehouse, and for governed company data it is usually the right place for it. With Paperswift the export never moves: it is opened inside your browser, and the only thing used when you ask a question is your column names and whether each holds text, numbers, or dates — never the values in any row. The privacy page sets the boundary out in full.

Which one should you choose?

Choose Snowflake, with Cortex on top, when an organisation needs governed data at scale and many people asking questions of it continuously. The loading, modelling, and consumption cost all buy something at that size.

Choose Paperswift when the unit of work is one export and one question, and standing up a warehouse to answer it would cost more in setup than the answer is worth. Most companies eventually need both.