The Automated Intelligence Manifesto: Six Promises About Your Numbers
Most of what is sold as artificial intelligence for data is a very good guesser. Automated Intelligence is a different idea: intelligence that is automated, in the sense that the work of reasoning through your data is done for you, step by step, in a way you can follow. These are the six promises we think a tool should make about the numbers you rely on.
Trinetro Labs · Published 12 October 2026 · 3 min read
1. The same question gets the same answer
Ask on Monday and ask again on Friday: if the data has not changed, the number has not changed. A figure you cannot reproduce is an opinion. Every answer in Trinetro Labs is deterministic, which is the quiet foundation under everything else on this list.
2. You can always see the work
A number without its working is a number you have to take on trust. Each answer carries a reasoning trace: the table it came from, the columns it used, how tables were joined and what was filtered. Your finance lead, your auditor and a new hire can all read it.
3. When it does not know, it says so
A confident wrong answer is worse than no answer. If a question is ambiguous, Trinetro Labs asks which meaning you intend. If "last quarter" cannot be applied to the data, it tells you, instead of returning an all-time total under a quarterly heading.
4. Your definitions belong to your team
What counts as revenue, as an active customer, as a closed deal: these are decisions your business has made. Governed team memory keeps them once. A person can keep their own, and an admin can promote one to a company default and withdraw it later, so everyone who asks gets the same meaning.
5. Your data stays yours
Access starts closed. A new member sees nothing until an admin grants it, and a grant can be as narrow as one sheet of a workbook. When the last person signs out, the connections and access grants for the workspace are cleared.
6. Intelligence should be automated, not artificial
The word to hold on to is automated. The profiling of your tables, the discovery of how they join, the choice of the right grain for a total: all of that work is done for you, the way an experienced analyst would do it, and shown to you. Where a language model has a place, on paid plans, it is to put an already-computed insight into plain English. It does not compute the answer.
What this does not mean
- It does not mean language models are bad. They are excellent at drafting, explaining and exploring. The promise is about the number, not the conversation around it.
- It does not mean the work disappears. You still decide what to ask and what to do with the answer. The tool takes on the part that should never have needed your attention.
- It does not mean every question has a single answer. When it depends on a definition, the tool asks, and remembers what you decided.
The terms used here are defined in the glossary. For the engine behind these promises, see the architecture, and for how access works, the security page.
Frequently asked questions
What is Automated Intelligence?
A way of answering questions about data by reasoning through its real structure instead of predicting a likely answer. The same question on the same data gives the same answer, and every step can be shown.
How is Automated Intelligence different from AI?
Most AI tools for data generate an answer from patterns, so it can vary and can be wrong with no warning. Automated Intelligence works the answer out from the data and shows how.
Does Trinetro Labs use a language model?
Not to compute an answer. On paid plans one is used to put an already-computed insight into plain English.
Why does it matter that answers are repeatable?
Because a number you cannot reproduce cannot be audited, reconciled or defended. Repeatability is what turns a figure into something a team can rely on.