The AI case is gathering momentum. The evidence of value is not.
We separate the commercial question from the technology story and write down what would actually justify the investment.
Clarity before commitment
Norelmarch works with leaders before a major digital or AI decision hardens into architecture, budget and organisational momentum.
The useful moment comes before the brief hardens
We separate the commercial question from the technology story and write down what would actually justify the investment.
We turn ambition into observable thresholds, failure states and a release decision that can be defended later.
We expose dependencies, incentives and alternatives before prior commitment becomes the reason to continue.
Three decision records
Each instrument answers one operational question: which option holds, when a pilot may ship and how far the evidence supports the conclusion.
Compares competing options across fixed constraints and makes sensitivity to changed assumptions visible.
Links evidence, validation and release states so a pilot cannot quietly redefine success after the fact.
Tests whether a verdict is supported by the supplied material, while keeping quotation, confidence and conclusion separate.
One decision, four working moves
The final document is meant to be used when Norelmarch is no longer in the room. It says what to do, why and what would change the answer.
Name the decision, accountable owner, admissible evidence and explicit exclusions.
Expose assumptions, contradictions, incentives, dependencies and missing proof.
Apply alternatives, red-team pressure and predetermined stopping conditions.
Recommend one justified move with uncertainty and limitations left attached.
The Norelmarch standard
Evidence can be traced back to the sentence, assumption or constraint that supports the conclusion.
Success and stopping conditions are agreed before momentum can redefine them.
The recommendation names the accountable person, immediate action and next review point.
The record stays with the recommendation.
Read the complete standardThree lines are enough to begin
Can UnalFounder · AI Systems ScientistEach review is led by Can from the first framing through the written recommendation. Send a concise note; no deck is required.