The Polarity Model
We read the relationship,
not the person.
Every mainstream tool measures individuals and adds them up. The Polarity Model takes the coupling between two people, or two functions, as the unit of analysis — because that is where work actually gets stuck.
The problem with adding people up
A team of excellent people can reliably produce a poor result.
Measure each person and you will find nothing wrong. Every individual is competent, well-intentioned and trying. And the organisation keeps producing the same outcome, year after year, absorbing every initiative aimed at fixing it.
That is not a people problem being measured badly. It is a structural property of the relationships between them — and it is invisible to any instrument whose unit of analysis is a person.
The scores move. The summaries pile up. The hours get counted. Nothing explains why.
What we look at instead
- The coupling between two agents — people, teams or functions
- Its state: productive, stuck or damaging
- The structural mechanism holding it there — authority mismatch, information asymmetry, a consent violation, an observer failure, a suppressed correction
- Whether the same pattern recurs in other rooms
The three states
The same read at every scale — and always a path out.
A single meeting, a relationship between two functions, or a whole organisation. One vocabulary, applied consistently, is what makes the ladder from one meeting to an organisational diagnostic a single instrument rather than three products. And at every scale the state comes with the coaching — the named structural change that moves it toward productive.
The exchange produced something. It moved, and it added more than it consumed.
It ran and produced nothing. The energy went in; nothing came out. Nobody is at fault and nothing is broken — the structure simply cannot close.
It consumed more than it created, and it suppressed the correction that would have fixed it.
Six dimensions
What the organisational diagnostic reads.
Purpose Integrity
Whether your organisation's stated purpose is genuinely operative — or whether a different, unstated purpose is actually driving decisions.
Consent & Commitment
Whether the agreements holding your organisation together are genuine, or made under pressure, habit or fear — producing compliance rather than real commitment.
Freedom-to-Operate
Whether the people in your organisation have the authority, information and space they need to act on their genuine judgement — or whether structural constraints are holding them below their capability.
Relational Health
Whether the key relationships in your organisation are genuinely collaborative — or merely transactional, or the hidden source of chronic friction.
Observer & Learning
Whether your organisation sees itself clearly and learns from what it sees — or whether it sees but does not act, or denies what it sees entirely.
Human-AI Integrity
Whether AI adoption in your organisation is happening with genuine human agency — or through drift, imposition or an unintentional dependency.
The sixth is unusual, and deliberately so. Human-AI Integrity treats AI adoption as a structural question about agency rather than an IT policy question — whether the adoption is happening with genuine human consent, or through drift, imposition and unintended dependency. We are not aware of another organisational diagnostic that carries it as a native dimension, though that is our reading of the market rather than a claim we can prove.
Published
Three papers, with DOIs.
The Polarity Model
A cross-scale structural framework for understanding agent, relational and systemic dynamics.
Relational Integrity in Multi-Agent AI Systems
A Polarity Model framework for human–AI safety.
Beyond Individual Agent Monitoring
Empirical evidence of three relational failure modes in LLM-based multi-agent systems. The empirical paper: what individual monitoring cannot see.
The next step is evaluation by people outside our own team. From autumn 2026, master’s students on strategic HR and leadership courses will run organisational diagnostics with the platform, take the moderator role, and critique the reports as coursework. Agreements are being finalised, and we will name the institutions when they are signed.
Design choices you can hold us to
The system is built to be able to say nothing.
An uncited finding is discarded. Every observation has to point at its evidence — a moment in a conversation, a question and a score, a document. Findings that cannot are dropped rather than softened.
Reports are withheld when the evidence will not carry them. Some meetings return nothing at all, and on the organisational side a meaningful share of processing has been spent on runs deliberately stopped before they produced a report. We would rather build a system that can decline than one that always has an answer.
The refusals show restraint, not accuracy. That distinction matters and we will not blur it. Whether what the system does say is right is answered by your own people recognising it, which is why validation before release is part of the method rather than a courtesy.
It describes conditions, never feelings. A rule inside the method forbids reporting how people feel. The output names structural conditions and real interfaces, which is also what keeps the product out of the employee-surveillance category.
Start here
Start with one meeting.
A permanent free account with three meetings to use in your first fourteen days. Unused ones expire at the end of the fourteen days; meetings you buy never do. No card. If the first report doesn't tell you something you didn't know, stop.
Google Meet and Microsoft Teams. Reports in English, Estonian, Finnish, Russian and more.