Soundings — diagnosis of complex human systems
An accurate diagnosis is the initial condition.
Everything a leader does afterward is path‑dependent on it. Get the diagnosis wrong and every good decision that follows is aimed at a system that isn’t there. Soundings is a diagnostic AI agent that reads an organization as the complex adaptive system it actually is — seven angles, nine stages, and findings that state in writing what would prove them false.
The shallow read
Almost everything called a diagnosis is a survey.
Engagement scores. A 360. An offsite where the leadership team writes on a wall what it thinks is wrong. An outside firm that interviews forty people and hands back the median of what they said. Every one of these asks the system about itself.
A complex adaptive system does not know why it does what it does. The behavior that matters is produced by structure the people inside it cannot see from where they stand: feedback loops that close somewhere else in the building, memory encoded in a procedure nobody remembers writing, incentives pointing somewhere nobody aimed them. Ask the system about itself and it will tell you about its intentions, sincerely and at length.
A survey collects what a system believes about itself. A diagnosis establishes what it is doing.
The cost of the difference is familiar. The obvious cause gets addressed. Consultants arrive. Something is restructured, an initiative launches — and six months later the problem is back, or it has moved somewhere else, or the fix created three new problems nobody anticipated. That is not a failure of effort.
The diagnosis
Soundings takes the depth.
The name is literal, and it is inherited.
At Tarawa in 1943 the landing craft grounded on a coral reef a mile and a half out that nobody had charted, and the Marines went into the water and waded the rest of the way in under fire. The Underwater Demolition Teams were created so that would not happen again — and the job was to go in first and find out how deep the water was and how the bottom sloped.
Not from a boat. In the water. A swimmer, at night or before first light, off an occupied beach, working a marked lead line in toward shore and taking the depth as he went. The line is marked so the man reading it always knows the difference between a depth he measured and one he estimated— and he reports them differently. Every time. In the dark, when a guess would be easier and nobody would ever know.
Soldiers and Marines wade ashore on that number. You do not round it.
That is the discipline this instrument is named for. It has carried it since its inception.
“By the mark”
The line landed on a marking. The evidence carries the finding — and the finding states the grade of evidence it rests on.
“By the deep”
The depth fell between marks. The evidence is too thin to carry a finding, so none is issued — and the silence is reported with its reason rather than quietly omitted.
“No bottom”
The line ran out before it touched. The record is opaque. Opacity is scored and reported as a result in its own right. What could not be seen never disappears into what was.
The reachable paths
Then it reads forward — never backward from where you would like to be.
A diagnosis that reasons back from a desired outcome is not a diagnosis. It is a proposal wearing one. Soundings reads forward from the diagnosed state and describes what is actually reachable from there.
The picture is a staircase. A destination is reachable if there are stable landings between here and there — intermediate forms the system can rest on without falling back — and the measurement that governs the risk is the longest stretch it has to cross with no landing at all. Which means a modest‑looking target with one wide gap in the middle is more dangerous than an ambitious one with good footholds the whole way up.
Sometimes a destination is simply not reachable on the current path. That is reported as a result, with the reason named, because knowing a door is closed is as useful as knowing one is open. And the instrument will not choose for you. Describing the ground is diagnosis; choosing the path is leadership. The two are held apart deliberately, because the moment an instrument starts preferring an outcome it starts finding evidence for it.
Complexity‑informed leadership
Knowing where you are does not move you.
The diagnosis names what is reachable. Walking one of those paths is leadership work — and in a complex system, leadership is not control. You cannot install a culture, cascade a behavior, or engineer an outcome three levels below you. Every organization that has tried produced the same result: compliance at the surface, and underneath it the original system, unchanged and now also cynical.
What a leader can do is shape the conditions under which the outcome they want becomes the likely one. The constraints that enable rather than restrict. Where feedback closes and how fast it returns. Who talks to whom, about what, with what consequence for being wrong. Slower to describe than a plan, considerably harder to fake, and the only thing that has ever worked on a system like this.
The warrior’s mindset
Going first is not the same as getting others to go.
For most of recorded history to lead meant to go before — to guide, to bring forth, to cross the threshold first. It was a discipline of the self before it was a claim on others. The Industrial Revolution recast the word as influence, and the first fight quietly dropped off the list. Nothing on an org chart requires it. Nothing measures whether it happened.
But designing conditions rather than issuing instructions asks something specific of a person: to act without a guarantee, to be corrected in public by someone junior and make that safe by going first, to sit inside not‑yet‑knowing long enough for the system to say something true. None of that is temperament. All of it is trained.
Judge it yourself
You should not have to take any of this on trust.
Everything above is a claim about an instrument, made by the person who built it — which is a poor position from which to argue that findings ought to state what would prove them wrong.
So a number of Soundings diagnoses have been written up and published, free to read: an adversary’s peace plan, a negotiation under scrutiny, an electricity grid, a technology transition read forward from present signals. Mostly systems that will not open their books to an analyst, which is the harder half of the problem. The reasoning is on the page. Go and disagree with it.
Or settle it more directly. The first diagnosis of your own system costs nothing — no fee, no obligation, yours either way. Details are on the instrument page.
The destination is a resilient organization.
One that absorbs a shock it did not forecast. That learns without needing a crisis to force the lesson. That does not depend on any one person being in the room, and does not quietly punish the people who bring bad news early.
You do not get there by wanting it. You get there by knowing precisely where you are standing, understanding which paths are genuinely open from there, and having a leader prepared to walk one.