Doram vs ColRows
Both build the model for you. They differ in what they read and how access rules work.
DoramChecked on 17 October 2026. Every fact about ColRows comes from its public website and docs. The links are at the end.
The short answer
ColRows is the closest tool to Doram on our list. Both build the business model for you, both send changes to people for review, and both run every answer through that model.
The differences are in the details. Doram also reads the tools where decisions are made, like Slack, Jira and the CRM. Doram follows your warehouse permissions, while ColRows keeps its own. And the two describe the path from question to SQL in different ways.
At a glance
| Question | ColRows | Doram |
|---|---|---|
| What it is | A semantic layer that builds itself, with an AI analyst | A context layer with its own AI analyst |
| Who builds the model | ColRows builds it. Curators approve or reject changes. | Doram builds it. Owners approve the changes that matter. |
| What it reads | Schema, column profiles, query history, Power BI and Superset metrics, catalogs, documents | Schema, column profiles, query history, dashboards, documents, Slack, Jira, CRM |
| How the AI gets the number | A compiler, or SQL drafted by the AI and checked before it runs | The AI picks approved facts. A semantic compiler writes the SQL. |
| Access rules | Its own rules, with masking. Off until you switch them on. | Reads your warehouse grants at every sync |
| Other AI tools | MCP, API, JDBC, Excel add-in | MCP and API |
| Where it runs | ColRows cloud, your private cloud, or fully air-gapped | Doram Cloud, or your own cloud account |
What ColRows does well
- The same rules on every surface. Row filters, column rules and masking apply in the chat, the SQL editor, the API, MCP and Excel. MCP calls run as the signed-in person.
- A model that keeps itself current. ColRows reads many signals, and when a schema changes, it marks every metric and saved query that depends on it as stale.
- Runs almost anywhere. It runs as a shared cloud, in your own VPC, or fully air-gapped with your own AI model. Tools like Tableau and Power BI can also read it over JDBC.
Where they differ
What each one reads
ColRows reads a wide set of sources: tables and keys, column profiles, query history, the metrics already defined in Power BI and Superset, catalogs like Unity Catalog, and documents you upload. Query history is off by default for each data source.
Doram reads the same kinds of data sources, and it also reads the tools where people make decisions: Slack, Jira, the CRM and billing. That is often where the reason behind a rule lives. Doram keeps that reason with the rule and shows it with the answer.
Access
ColRows has strong access rules, but they live inside ColRows. Each data source uses one stored login. Its “Restricted Access” switch is off by default, and while it is off, ColRows skips its own access checks and queries run with that login’s permissions.
Doram reads your warehouse grants at every sync, so each person sees only the rows they can see today. A revoke reaches Doram straight away. When two rules disagree, the stricter one wins.
How the SQL gets written
ColRows explains this in two ways. Its blog says the compiler writes the SQL. Its core docs say the AI may draft the SQL, and that the draft is checked for read-only access, permissions and cost before it runs. If this matters to you, ask them which path your questions take.
In Doram, the AI never writes the SQL. It picks approved facts, and the semantic compiler writes the query.
How changes are approved
In ColRows, curators review what the system proposes, and only accepted meaning is used for answers. We found no rule that every change must wait for a person.
Doram routes each change by risk. Small changes go live on their own, and changes to a number your team reports on always wait for the owner. Answers only use facts that are approved or certified.
Accuracy claims
ColRows reports 98.2% accuracy for its compiler, against 14.5% for a raw AI model. This is its own test, on a private set of 50 schemas plus two public medical datasets. The questions and the baseline model are not published. Doram does not publish an accuracy number yet.
Who should pick which
Pick ColRows if you need a fully air-gapped setup, your BI tools must read the model over JDBC, or you want to import metrics from Power BI and Superset.
Pick Doram if many of your rules live in Slack, Jira, the CRM and documents, and you want access to follow the grants you already set in your warehouse.