Method
How we work
Every engagement runs through the same five stages. Each step below says what it produces, what the machine does and what a person decides.
Gather
Primary documents first: FCC orders and licence data, SEC filings, auction results, standards contributions. Trade press is used, marked as reported, and never upgraded to fact.
AI does
Fetches, parses and cross-indexes at a scale a person cannot. Flags where two sources disagree.
A human always
Chooses the sources, and checks every figure against the original document before it is used.
Model
Arithmetic is shown, in one unit, with its basis stated. A per-market figure is never averaged into a national one without saying so.
AI does
Builds and runs the model, and tests it against edge cases.
A human always
Sets the assumptions, and owns the number that goes in front of your board.
Challenge
Every conclusion is attacked before you see it. The counter-case is written down, not implied. If we cannot state the strongest argument against our own recommendation, we do not understand it yet.
AI does
Red-teams the argument and hunts for the contradicting source.
A human always
Decides what survives, and signs their name to it.
Recommend
One recommendation, its counter-case, and what would have to be true to change it. Plus the observable indicators that will tell you which way it is going, so you are not dependent on us to find out.
AI does
Nothing. Drafting is not deciding.
A human always
Writes it.
Every claim carries its class
So you can tell, without asking, what kind of sentence you are reading.
Filed: a company’s own filing or a regulator’s order. Calculated: our arithmetic on filed figures. Reported: trade press, attributed and unverified. Judgement: our reading of the evidence.
THE CORRECTION PROMISE
If we get a figure wrong, the correction appears on the piece that carried it. Two so far: a satellite operator’s revenue we carried forward a year, and a transfer we described as pending when it had been approved.
It is published so that you can check it.
What a person decides
- Choosing which sources count
- Setting the assumptions in a model you will act on
- Deciding what the evidence means
- Signing a recommendation
AI is what lets one person do the work of a team on public data. The judgement is ours.