Columns AI
Columns AI is an end-to-end data platform built around Drive, Flow and Graph. Spreadsheet-bound teams extract data from files, transform it in plain English, then publish live visual stories without writing SQL or code.
What is Columns AI?
Columns AI is an end-to-end data platform that promises to take a team from raw files to published visuals inside a single product. It is organised around three components. Columns Drive behaves like a cloud drive, but instead of merely storing files it runs OCR and AI on every .png, .jpg, .pdf or .txt dropped into it and returns structured data: product screenshots, invoices, receipts and bank statements are the documented targets. Extraction is contextual, so the folder name and the files already processed help keep the schema consistent for new arrivals, and a Drive folder can then be reused as a live data source.
Columns Flow is the transformation layer: a visual canvas where each node is a dataset and each edge a transformation. Rather than writing SQL, the user states an intent in one sentence, and Flow generates the computation plan, previews the resulting sample, and applies it only after approval. Multi-source operations cover Union, Join, Lookup and Except. Columns Graph turns the cleaned data into visuals through a query engine of metrics, dimensions and filters, with pivots and comparison, percentage or cumulative transformations. Eleven chart families are documented, from maps and timelines to bar races, radars, box plots and word clouds, and because the canvas is SVG almost every element can be clicked and restyled directly. A full-context AI chat answers questions, suggests charts and generates styles from a few keywords.
Sources include CSV/TSV, Drive, Google Sheets, HTTP APIs, Notion, Airtable, Postgres and Snowflake, with MySQL and BigQuery still announced as upcoming and an embedded SQL IDE for warehouses. Output goes out as a public link, an iframe embed with domain whitelisting, PNG/SVG/CSV exports, a Slack share or a presentation deck, and a visualization API lets paid accounts replay a template with new data. Team features include SAML SSO and multi-seat codes. The company, Columns AI, Inc., was founded in 2022 by Shawn Cao and claims more than 12,000 professionals.
What it does
- Extract structured data automatically from PDFs, PNG, JPG and TXT files through OCR and AI, with the folder context improving schema consistency
- Build a visual node-based pipeline by describing each transformation in plain English, previewing the plan and sample output before approving it
- Connect Google Sheets, Excel/CSV, Notion, Airtable, HTTP APIs, Postgres or Snowflake as live data sources
- Query data in natural language and get a chart back, then restyle every element on an editable SVG canvas
- Combine several sources with Union, Join, Lookup and Except operations
- Publish by link, embed as an iframe, export to PNG, SVG or CSV, share to Slack or assemble an interactive presentation deck
- Schedule hourly, daily or weekly refreshes and fire Slack, email or webhook alerts when a data rule triggers
When to use Columns AI / When not to
A quick filter to help you decide if Columns AI is the right fit.
When to use Columns AI
- Business, marketing and RevOps operators who live in spreadsheets and need recurring dashboards without waiting on a data team
- Data and business analysts who want metrics, dimensions, pivots and one-click forecasts without standing up a full BI stack
- Consultants and freelancers delivering client-ready reports, presentation decks and embedded charts that keep updating on their own
- Finance, accounting and back-office teams turning invoices, receipts, purchase orders and bank statements into structured tables through OCR extraction
- Media professionals, bloggers and business leaders who embed self-updating charts in articles, sites and KPI reviews
When not to use Columns AI
- Buyers who need a formal compliance file: no SOC 2, ISO 27001 or HIPAA certification is claimed, no DPA is offered and the GDPR is never mentioned
- Teams looking for general-purpose file storage, since the documentation states plainly that Columns Drive is not a general-purpose storage tool
- Anyone who needs a mobile app: there is no iOS or Android application, only a Gmail add-on and a Slack app
- Organisations requiring on-premise deployment or a chosen data residency, as every analysis runs in the vendor's own cloud
- Teams standardised on MySQL or BigQuery, both still announced as upcoming, or needing an interface in a language other than English or Japanese
How to use Columns AI
A typical end-to-end flow, from setup to results.
- Create an account with Google, Slack, SSO, or an email and password — the Gmail add-on even provisions one automatically
- Connect a data source: Google Sheets, Notion, Airtable, an HTTP API, Postgres, Snowflake, a CSV/TSV upload, or a Columns Drive folder
- For files, drop them into Drive and let the extraction run: allow up to five minutes per file when it arrives through the Gmail add-on
- In Flow, start from the input node and sync it to load the schema and sample rows
- Add a node, link it to the input, then open the edge detail to describe the transformation you want
- Write the intent in one sentence and click Generate — for example convert the date field to month value only, like yyyy-mm format — reusing the exact column names shown in the schema
- Review the generated plan and the output sample, then click Approve to update the target node
- Move to the analysis step: let the AI chat suggest charts, or build the query yourself with metrics, dimensions, rules and a timeline
- Style the result on the SVG canvas with annotations, images, shapes, arrows, widgets, typography and palettes, saving defaults or generating a style from keywords
- Share it: publish a link, embed an iframe, post to Slack, export, or build a deck — then use the story menu's Schedule option to keep it refreshed
Pros & Cons
Pros
- The whole chain sits in one product: raw file, clean data, pipeline, published visual — with much less tool switching than a classic BI stack
- Transformations are described in plain English, with no SQL or code required, and previewed before they are applied
- OCR extraction from PDFs and images lives in the same tool as the analysis, so invoices and statements become datasets without a separate step
- The SVG canvas is unusually editable, allowing graphic customisation well beyond what a standard BI tool permits, and distribution is well thought out: live embeds, domain whitelisting, decks and multiple exports
- A free tier with no time limit, plus USD 10 of credit for everyone, means the tool can be tested without commitment or a sales call — prices are published openly
- Full free access is announced for teachers and students, with discounts for non-profits
- Public documentation is complete, backed by fifteen video walkthroughs, a visualization API with sub-second rendering, and a Gmail add-on and Slack app for everyday workflows
Cons
- No postal address, no detailed legal entity and no registration number are published anywhere on the site
- No GDPR mention, no DPA, no Article 27 representative and no security certification (SOC 2, ISO 27001, HIPAA)
- The privacy policy carries no effective date, and the terms state they were created with the help of the Terms Of Service Generator, naming the laws of us as governing law with no identifiable jurisdiction
- Personal data retention is announced as 90 days maximum, with no stated duration for project data or uploaded files
- The site is entirely rendered in JavaScript, so pricing, legal notices and the contact page are invisible without a browser, search engines included — and two different pricing grids coexist in the shipped code, which can confuse
- No mobile application, an interface limited to English and Japanese, and MySQL and BigQuery connectors still announced as upcoming
- Billing is per processed row: the cost follows the volume and is steered by credits rather than by a configurable spending cap, and no training opt-out is documented since the vendor states no training takes place
Pricing & Plans
Columns AI offers a permanently free plan. The Basic tier costs USD 0 per month with no time limit, usage being billed at USD 0.05 per 1,000 rows processed, and every new account receives USD 10 of credit on sign-up. The lowest paid entry point is the Premium plan at USD 20 per month, which reduces the unit rate to USD 0.03 per 1,000 rows; Platinum follows at USD 50 per month at USD 0.01 per 1,000 rows. An Enterprise offer is available on quotation via support@columns.ai. There is no free trial, and payment and billing are handled through Stripe.
- all three products (Flow
- Drive and Graph)
- unlimited creation
- pay-as-you-go usage and community support
- everything in Basic
- a 40% lower per-row rate and priority email support
- the vendor positions it as worthwhile above 1 million rows per month
- everything in Premium
- a 60% lower per-row rate and Elite account support
- positioned as worthwhile above 2 million rows per month
- custom pricing
- dedicated support and advanced controls
- through the Talk to sales route at support@columns.ai
- Education — free for teachers and students
- by signing in with a .edu address or sending proof of status to support@columns.ai
- the visualization API is reserved for paid accounts
- and plans can be changed at any time
- usage then being billed at the current unit rate
Data, GDPR & hosting
A consolidated view of how Columns AI handles your data.
GDPR overview
The site never mentions the GDPR: the word appears nowhere on it. What exists instead is a Your Rights section in the privacy policy addressed to European residents, offering access, correction, updating and deletion of personal information, on a stated basis of contract performance or the company's legitimate interests. The same policy warns that information will be transferred outside Europe, including to Canada and the United States, without naming any transfer mechanism: no standard contractual clauses, no Data Privacy Framework. There is no Article 27 representative, no named DPO and no Data Processing Agreement on offer. Deletion, access and portability requests go by email to support@columns.ai. The policy carries no effective or last-updated date, and the terms name the laws of us as governing law without identifying a jurisdiction.
Who owns the data?
The terms grant each user complete ownership of the content they create on Columns AI, including downloading, embeding and distributing. Storage is split: uploaded CSV/TSV files and documents dropped into Drive are kept in Columns' cloud storage, while connected sources such as Google Sheets, Notion, Airtable, Postgres, Snowflake and HTTP APIs stay in the user's own systems and are synchronised on demand. The FAQ states that Columns does not store your data apart from file uploads, everything else being processed at runtime. Google account details sit in a Firebase profile, and all stored personal data is removed on an account deactivation request, available from Settings. One caveat: files added to Drive get a unique public link by default, which can be switched off.
Reuse rights
Users may reuse what they build without asking permission: the terms allow them to download, embed and distribute their own content freely. In the other direction, Columns AI automatically collects device and browsing information — browser, IP address, time zone, cookies, log files, web beacons, tags and pixels, pages viewed and referring search terms — plus the email address from the public profile when a purchase is made. The stated purpose is to identify the user and smooth navigation between Columns and other cloud applications such as Google Sheets; the policy says this information is not used for any other purpose, is not shared with third parties, and that Do Not Track signals are not honoured. AI processing is delegated to OpenAI's API services and Google Cloud Vertex AI over encrypted APIs, in real time only, sharing the minimum necessary context or data required to fulfill your specific AI-driven request, and the policy states this data is not used by those providers to train their public models. The Google Sheets connection is read-only and revocable from the profile. Google Tag Manager, Google Ads, Microsoft Clarity and Chatwoot trackers are present on the home page.
Data retention & training
Hosting summary
Columns AI does not publish a hosting country as such. The domain resolves to 35.241.5.23, an anycast node on AS396982 (Google LLC, Google Cloud) geolocated in the United States, and the privacy policy places user profiles in Google Firebase. Analyses run inside Columns' own cloud cluster. Uploaded CSV/TSV files are kept in Columns' cloud storage, while connected sources are not copied and are synchronised on demand. The policy warns that information will be transferred outside Europe, including to Canada and the United States, which is a transfer destination rather than a declared hosting region: no data residency option is offered. Announced protections are TLS 1.2 or higher in transit, AES-256 encryption at rest for sensitive data such as database records and account credentials, least-privilege access control with MFA, and regular security audits. What is not said is equally relevant: no named data centre, no region choice, no sub-processor page, and no certification backing the stated measures.
Things to keep in mind
Risks and trade-offs to weigh before adopting Columns AI.
- Cost tracks volume: even the free tier bills USD 0.05 per 1,000 rows processed, and the only ceiling is the credit left on the account, not a spending cap you can set
- Files dropped into Drive receive a unique public link by default — turn it off before uploading invoices, statements or anything else confidential
- No security certification, no DPA and no GDPR mention: check this carefully before processing European personal data, and remember transfers to Canada and the United States are announced with no transfer mechanism named
- Automatically generated terms with governing law stated as the laws of us, and a privacy policy carrying no date, make it impossible to know what applies and since when
- Two pricing grids coexist in the site's shipped code; the one served on the pricing page is the current one, and any older annual grid with a seven-day trial should be ignored
- Public link sharing is powerful but exposing: check domain whitelisting and password protection before circulating a chart built on sensitive figures
- Plain-English transformations can hide what actually happened to the data — read the generated plan and the sample output before approving, rather than trusting the preview at a glance
Setup & Integrations
Technical difficulty
Low. Sign-up takes one click through Google, Slack or SSO, the Gmail add-on even creates the account by itself, and there is nothing to install, deploy or maintain — it is a web application. Connecting a source means OAuth for Google Sheets, Notion and Airtable, or credentials for Postgres and Snowflake. No SQL is needed on the main path, though an SQL IDE is available for those who want it. The only real learning curve is phrasing: the documentation advises naming columns exactly as they appear in the schema. Programmatic use requires generating an API key in Settings.
Deployment
Integrations
Supported languages
Behind Columns AI
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Columns AI.
Frequently asked questions
What is Columns AI?
Do I need to know how to code or write SQL?
Which data sources can I connect?
How much does it cost?
Is there a free trial?
What counts as a billed row?
Is my data used to train AI models?
Is my data stored?
Can I keep published charts up to date?
Is the GDPR covered?
Should you pick Columns AI?
Columns AI is a coherent end-to-end proposition for people who live in spreadsheets and have to produce visuals that stay up to date. The three products fit together honestly: Drive converts files into data, Flow cleans and reshapes it from a plain sentence, and Graph turns the result into a chart that can be published, embedded, scheduled and restyled down to the last SVG element. For an operations team, a consultant or a freelancer, this removes several tools at once, and the commercial terms make it easy to evaluate: prices are public, the Basic tier is free with no time limit, USD 10 of credit is granted on sign-up, and teachers and students get the platform free. There is no sales gate and no free trial to negotiate — you simply start.
The weakness is elsewhere, and it is clear. Legal transparency and compliance are thin: no postal address anywhere on the site, no named legal entity beyond Columns AI, Inc., no GDPR mention at all, no Data Processing Agreement, no Article 27 representative and no security certification. The terms were produced by a generator and name governing law as the laws of us without a jurisdiction, and the privacy policy carries no date. The publisher is young and small, with a single founder in the shop window, and its billing follows the volume of rows processed rather than a spending cap you can set.
The verdict follows from that split. For an operations team, an independent professional, a newsroom or a classroom, Columns AI is a genuinely good choice, and the free tier costs nothing to evaluate. For a regulated environment or a procurement process that requires a compliance file, it needs proper due diligence first — and, on today's published information, the answers may simply not exist.
- Choosing a selection results in a full page refresh.
- Opens in a new window.