Paper III · The series
The firm sees process. The trader keeps the conversation.
Supervision needs records. Coaching needs candor. Put both in one channel and you lose both. This paper reads the off channel enforcement sweep as a finding about human behavior, and describes the two plane design that lets a firm keep a better record than surveillance produces while its traders keep a place to think.
One channel, two failures
Supervision needs records: durable, attributable, reviewable by people with authority over the person recorded. Coaching needs candor: the freedom to say “I want to double down because I am angry” to someone, without the sentence becoming an exhibit. Force both into one channel and the record fills with performance instead of truth, while the coaching starves, because nobody confides in a microphone wired to their supervisor.
The industry has run this experiment at scale. The off channel communications sweep, billions in penalties since 2022, is usually told as a compliance failure, and it is one. Read from the other side, it is also a finding about behavior: when every sanctioned channel is a supervised channel, people take their real conversations elsewhere. Surveillance did not capture the candor. It relocated it.
So the design question for a record at the intent stage was never how to record more. It was how to record the right things while leaving one place where the truth can be spoken. The answer is two planes with a wall between them, and the wall is the product.
The two planes
The trader plane carries the coaching dialogue: the session with DJ, the doubts, the reasoning, the bad first drafts of a plan. It is never summarized, scored, or forwarded, and never reachable from any firm facing surface. This is not a permission an administrator could flip. It is structural, and the deployment’s own plane denial report demonstrates it, in the firm’s tenant, for the firm’s security team to verify.
The firm plane carries what supervision needs: committed plans, mandate evaluations, observations with figures, briefs the trader chooses to send, acknowledgements, attestations, the audit ledger. When a trader commits a plan, the plan and its evaluation cross the boundary, timestamped. Nothing else does. Written as a rule for the trader, and we put it in writing before any pilot session: the firm sees process, you keep the conversation.
What each side gets
The trader gets a coach that can hear the truth. And clause by clause, the firm decides how much of the measured figure the trader sees. Some clauses run full visibility. Others run direction only: the trader is told the committed plan sits past a limit; the exact figure stays with compliance. The trader learns what they need in order to act. The firm keeps the measurement authority. Both facts are printed where each party can see them, so nobody discovers the arrangement later.
The firm gets a better record than surveillance ever yields. Surveillance harvests fragments and infers intent. Here, intent arrives structured at the moment of commitment: instrument, size, levels, thesis, evaluated against the mandate’s own clause text, with the projected figure, seconds after the trader confirms. It is a smaller record than a transcript and a far stronger one, because every line is a decision the trader deliberately put on the record. There is a quieter benefit too: reasoning written in the knowledge that it will cross is the trader thinking clearly on purpose. The commit is a small act of self supervision before any officer reads a word.
Why the wall cannot be a setting
Every enterprise instinct says make it configurable; let the firm choose its privacy level. We refused, for one reason. A boundary management can move is, in the trader’s mind, already moved. The first week’s candor would be the last, and every session after it would be performed for an imagined future reader. The value of the trader plane rests entirely on the trader believing the wall, and traders are professionally skeptical people. They believe architecture, verified by their own firm’s security team. They do not believe toggles.
That is also why the product states its limits out loud on the firm side. The register says which clauses could not be measured and why; the trader’s session says which clauses could not be checked. A system this honest about what it cannot see earns the right to be believed about what it refuses to see.
The trade both sides make
The firm gives up the fantasy of total capture and receives a decision record of a quality surveillance cannot produce, populated voluntarily, at the only stage where a record can still change the outcome. The trader gives up the fantasy of leaving no trace and receives a private place to think and a coach worth confiding in. Candor and evidence coexist. They just cannot share a channel. The wall between them has to be real, verifiable, and permanent, and then both sides can lean on it. It is why traders and risk officers can want the same product for different reasons. The final paper is about the other line we drew in permanent ink: the system’s relationship to the order itself.
Traders believe architecture, verified by their own firm's security team. They do not believe toggles.