Pre-commitment review
A strong analysis of the wrong problem is wasted work
And a good decision can become a bad one if the world changes and nobody notices.
AGI²⁺ is a human-sovereign institutional intelligence architecture for consequential decisions over time. Its first operational application is the pre-commitment review: a structured pass over a strategic decision before the commitment is made. It helps leadership identify whether it is solving the right problem, what it does not yet know, and under what conditions its intended decision would be wrong.
Two problems, arriving at different times. Before the commitment: are we solving the right problem? After it: are the assumptions and mechanisms behind that decision still true? The review is built for the first. The runtime carries the second. Neither produces the answer; the people who hold the authority decide.
The Decision Loop
Order matters, because each answer changes how the next question is read
-
Frame
The real problem, and what has been left outside it.
-
Hypotheses
The preferred explanation and its strongest rival.
-
Question
The one question whose answer would most change the picture.
-
Evidence
Where it came from, and whether it is independent.
-
Belief
The revised position, contradictions kept visible.
-
Readiness
Whether another week of analysis can still change the outcome.
-
Decision
Taken by the people who hold the authority.
-
Action
What is authorised, and inside which limits.
-
Outcome
What happened, recorded against what had been assumed at the time.
-
Learning
What the organisation now knows.
After the decision
The Decision Loop stops at Learning. The organisation does not.
A second loop runs after the commitment is made: the Persistent Institutional Loop. It carries the commitments, deadlines, expected observations and falsifiers forward, watches what actually happens, checks whether the mechanism the decision relied on still operates, and re-anchors the state the next cycle starts from. That is what AGI²⁺ v26.0 adds, and it is architecture rather than shipped software.
What the review produces
The four things a review gives you
Frame check. Are you about to answer the right question, and whose objectives does that question actually describe?
Crux list. Which two or three uncertainties decide this, and which of your open workstreams cannot change the outcome?
Decision structure. Is this a one-way door? If it is, what would turn it into a two-way door — stage the commitment, or write the exit trigger before you sign?
Signposts. For each crux, which observable indicator would show the assumption degrading, and at what threshold do you act?
Modes
Five modes, one runtime
Learning
Optimises for what the person can still do once the assistance is withdrawn.
Research
The product is a defensible claim, and primary sources come before commentary.
Strategy
Market entry and competitive response. The visible output stays a question until readiness.
Board
Adds legitimate authority and decision provenance, so the record holds what was known at the time and what was only assumed.
Adversarial / Cyber
Assumes that some of what you are looking at has been manipulated on purpose.
What you will not get
You will not see probability charts or forms on screen. The mathematics stays behind the interface, and there is no score at the end. The runtime also keeps no record of who was wrong, and that is deliberate: organisations have shut down internal forecasting tools that beat their own experts, because the tools created a record of who had been mistaken, and no leadership team buys that twice.
Who decides
The machine may analyse, challenge, monitor, test and propose. Authority over purpose, values, commitments and acceptable risk stays with the people who carry the consequences. In AGI²⁺ that is an architectural constraint, written into what the runtime is allowed to do, and it holds in every mode.