Every Counterparty Is a Human System

A company cannot want anything. It cannot be short of money in a particular month, cannot owe a favour to someone who helped it thirty years ago, and cannot be six weeks from an unannounced retirement. Each of those descriptions belongs to a person, and each has at some point been the thing that decided the matter.

The entity is a legal convenience, and it is the unit around which nearly all commercial diligence is organised: the file is opened on the company, the searches run against the company, the risk written up as a property of the company, as though a corporate form could be reliable or unreliable the way a person can. The framing is not false, only displaced: one level above the layer where behaviour happens.

This firm states the principle on its investigations page in a single sentence: every company, counterparty, and risk is ultimately a human system, and reading the people, their incentives, loyalties and pressures, is what turns raw information into decision-grade intelligence. What follows is the argument for it, and the argument against the objection it invites, which is that reading people is a polite word for guessing about them.

The law has already moved

On 29 June 2026 section 250 of the Crime and Policing Act 2026 came into force. Home Office Circular 004/2026 sets out the effect: where a senior manager of a body corporate commits a criminal offence while acting within the scope of their actual or apparent authority, the body corporate commits it too. The equivalent provision in the Economic Crime and Corporate Transparency Act 2023 was confined to specified economic crimes. This one is not confined by offence at all. The circular is specific that the test turns on a senior manager's actual roles and responsibilities, not their title.

As policy that widens corporate exposure. As a statement about where risk lives it does more. The older identification doctrine looked for a directing mind: one person, near the top, whose intent could stand in for the company's. The statutory test replaces the search for a mind with an inventory of people defined by function. To know what an organisation is exposed to you must now know which individuals can bind it by their conduct, and something about each. No diligence organised at entity level produces that list. It was never looking at that layer.

Incentives: what does this person gain, and over what horizon

The standard question is whether the counterparty is solvent and whether the deal makes commercial sense for it. Both are answerable from documents, and both can be answered yes while the arrangement fails, because an entity's interest and the interests of the people executing it run on different clocks.

Governance codes concede the point. Provision 36 of the UK Corporate Governance Code, applying to financial years beginning on or after 1 January 2025, asks that executive share awards vest on a phased basis rather than all at once, the Financial Reporting Council's stated intention being to hold attention on long-term performance. Such a rule is unnecessary unless a person's payoff clock and an enterprise's clock are known to be different instruments.

The reading it produces is chronological rather than moral. When does this individual's horizon end, and does it end before or after the horizon you are underwriting? An earn-out, a fixed-term secondment, a fund near the end of its life, a retirement decided but not announced. A counterparty can be entirely straight and still be optimising over eighteen months while you are committed for ten years. That is not misconduct but a fact about time, and it tells you which positions will be defended when the two of you diverge.

Loyalties: to whom is the obligation owed when two conflict

Section 175(1) of the Companies Act 2006 is unusually candid. "A director of a company must avoid a situation in which he has, or can have, a direct or indirect interest that conflicts, or possibly may conflict, with the interests of the company." The statute treats conflict as ordinary rather than exceptional, and the machinery built around it is a disclosure regime. Disclosure establishes that a conflict exists; it cannot establish which way it resolves.

Loyalty is plural, it is ranked, and the ranking is almost never written down. Obligations run to a family, to a first employer, to the person who made the introduction that made a career, to a community that outlasts the transaction. The moment that matters is the one in which two are live at once.

So the question is not whether an interest has been declared. It is which obligation was paid first, before, when the person had to choose. That is a matter of record: prior roles, conduct in them, what happened at the last point of collision. The firm's deep-level diligence names the object plainly: the people behind the target, and their past behaviour in other roles. It is a harder question than the declaration question and a more useful one.

Pressures: what is happening to them that has nothing to do with you

The Association of Certified Fraud Examiners published Occupational Fraud 2026: A Report to the Nations on 12 May 2026, drawn from 2,402 cases across 143 countries. Three findings bear on this.

Eighty-four per cent of perpetrators displayed at least one behavioural red flag before detection, and the flags recorded are overwhelmingly circumstantial rather than dispositional: financial trouble, living beyond means, comparable strains. They are facts about a situation, not a character, and are visible to people nearby long before they surface in any record.

Median losses caused by owners and executives were more than nine times those caused by employees. Risk is not evenly spread across an organisation; it concentrates where authority concentrates, which is where the identification doctrine arrived from the other direction.

Forty-three per cent of the frauds were detected after a tip, more than half of them from employees. The most productive detection channel in the study is a person who noticed something and said so.

The third reading therefore asks what is happening to this individual, or their institution, in the period that concerns you and has nothing to do with you. A restructuring. An unsettled succession. An exposure that makes an ordinary decision expensive. Pressure does not predict conduct. It tells you where to look, and it is the reading most often left out, because it requires talking to people rather than pulling records.

Why a behavioural bench sits inside an investigations house

Human terrain work asks who actually decides in a place, what they owe, to whom they answer, and what they fear. Commercial diligence on a joint-venture partner asks the same four questions of a company. The material changes; the question does not. Mapping a municipal power structure and mapping the real decision rights inside a group are one analytical task on different ground, which is why the anthropologists and the investigators are not in separate buildings.

Sources are themselves human systems, which is the second and less obvious reason. A former colleague willing to speak has a reason for speaking, and the reason is part of the evidence. Source validation, the weighing of access, distance from the event and interest in the outcome, is a behavioural discipline before it is an investigative one. A house that collects from people and does not keep the science of assessing people runs its most valuable channel unaudited.

What the method cannot do

The largest meta-analysis of deception judgements, published by Bond and DePaulo in Personality and Social Psychology Review in 2006, pooled 206 documents and 24,483 judges and found average accuracy of 54 per cent: 47 per cent of lies correctly identified, 61 per cent of truths. Unaided human judgement about whether another person is lying is barely better than a coin. That is the evidence base, and why this firm confines its deception work to scientifically defensible methods and says so about the alternatives: body-language frameworks and neuro-linguistic material sold with far more confidence than the literature will bear.

The consequence for a client is a refusal. We will not certify that a person is honest, and will not deliver a verdict on anyone's character, because no method available to this firm or anyone else supports one. What the three readings produce is narrower: a ranked set of possibilities about a named position, each carrying a probability term and the observation that would change it. The National Cyber Security Centre states that it applies the Professional Head of Intelligence Assessment probability yardstick to every judgement it makes. The discipline in writing that way is that the judgement can later be shown to have been wrong. Anyone offering more certainty than that about a human being is selling something the evidence does not carry.

Two further limits follow. Enquiry into people is conducted within the data protection and privacy law of each jurisdiction and only where a lawful basis exists, which rules out a portion of what any client might want. And a reading is taken at a moment: incentives expire, loyalties are re-ranked, pressures arrive. It has a shelf life and should be dated.

The statutes, codes and studies here are public instruments, set out so a reader can check them. They are not advice on any transaction, and carry no representation about any person, company or matter.

The principle is applied commercially through Corporate Intelligence & Investigations Group and Due Diligence Investigations; the science it rests on sits in Multilateral Human Terrain Mapping and Deception Detection, Truth Elicitation & Human Behavior.