A decision with no visible line
The system acts, and nobody can say afterward which cases it handles alone versus which should have gone to a person.
TEZZ · HOW IT DECIDES
A stated line between routine and judgment
How an AI system decides means which tasks it handles on its own, which it flags for a person, and where that line actually sits. A system that cannot describe that line is a black box, regardless of what it is built on.
TEZZ is scoped to routine, repeatable tasks: drafting, prioritising, flagging. Anything ambiguous or judgment-heavy is routed to a person rather than actioned automatically.
THE RANGE OF PRACTICE
What a black box looks like, and what a scoped line looks like
Described by behaviour, not by a score.
Either fully manual or fully automated
No middle ground between a person doing everything and a system doing everything, with nothing in between reviewed.
A stated line between routine and judgment
Routine, repeatable tasks handled directly. Anything ambiguous routed to a person before it happens, not after.
WHERE WE SIT
Minimum human input, not no human input
One of the engineering principles behind every system we build.
Minimum human input means removing repetitive manual work, not removing judgment from decisions that need it. TEZZ handles routine, repeatable tasks directly and routes anything that changes something materially, or that a reasonable person would want to weigh in on, to a human first. It is not deployed as a standalone system, it runs inside the platforms we build and is reached only through their API.
STRAIGHT ANSWERS
Questions about how TEZZ decides
The ones that come up before anyone signs anything.
Which tasks it handles on its own, which it flags for a person, and where that line sits. A system that cannot describe that line is a black box, whatever it is built on. TEZZ is scoped to routine, repeatable tasks, with anything ambiguous or judgment-heavy routed to a person.
On routine, repeatable tasks, drafting and flagging, yes, consistent with minimum human input as an engineering principle. Anything that changes something materially, or that a reasonable person would want to weigh in on, is routed to a human rather than actioned automatically.
Minimum human input means removing repetitive manual work, not removing judgment from decisions that need it. The two get conflated often, and treating them as the same thing is how an AI system ends up making calls nobody actually reviewed.
Within a Gigagrams platform, decisions TEZZ makes are attributable to a specific action taken at a specific time, consistent with the audit logging principle applied across our systems generally.
Building on a Gigagrams platform and want TEZZ involved?
Tell us what you are building and we will scope what it can assist with, and where the line to a human sits for your specific workflow.