The Login Quotient

Why this is the first question, not a technology question

A login is a permission. A job is a commitment.

A login can be counted, and so it gets counted — seats issued, departments active, prompts submitted, hours reported as saved. Those numbers answer whether a capability has been distributed. They do not answer whether the organisation has created value, and they say nothing at all about who answers for the result when it is wrong.

Most companies give AI access and call the activity strategy. That is not an operating model. It is a login.

The uncomfortable version of the question is the one that goes quiet in the room: who is responsible for the result? Not who has access to the tool. Not which vendor supplied it. Who will explain the outcome to the customer, the board, the auditor or the regulator, and who has the authority to stop it.

The Login Quotient measures the distance between those two things — what you have distributed, and what you have actually assigned.

What the number deliberately excludes

The denominator is consequential recurring workflows, not all AI use. A company should not build a bureaucracy around every prompt, and an instrument that demanded a charter for each one would be wrong about how organisations work. Reversible, low-stakes, internal drafting is meant to stay ungoverned. That exclusion is what makes a low score meaningful rather than merely strict.

What it cannot tell you

Coverage is not effectiveness. A charter can exist and still be ignored; a reviewer can be named and have no time, no competence or no authority to reject. This is Gate 0 of a longer instrument, and it measures whether the work has been assigned — not whether the assignment holds under load. The gate that tests that comes later, and it is arithmetic rather than opinion: divide the reviewer's genuinely available hours by the hours of checking the role actually requires. Below one, the reviewer cannot keep up and begins waving work through. At that point the control exists on the org chart and nowhere else.

The E-mployer's Instrument

The Login Quotient is the front door. Behind it is a deterministic sequence that takes a single candidate AI role from "someone suggested this" to a signed commissioning decision — a defined control level, an explicit authority boundary, a named reviewer whose capacity has been checked, an honest economic case, and written conditions under which the role stops.

It is not a maturity score, and the refusal is deliberate. Scoring models are compensatory: a strong number in one column quietly buys forgiveness in another, so a compelling business case papers over the fact that nobody is checking the work. The Instrument does not permit that trade. Some questions are pass/fail, and no amount of good news elsewhere buys them back.

1. The Login QuotientOrganisation-wide. Where you are now. (This page.)
2. The Necessity GateSix conditions a candidate must satisfy to be a role at all. Any failure stops the process — and which one fails determines what you do next.
3. Consequence ClassThe worst harm dimension governs, never the average. This sets how much human control is mandatory.
4. Oversight FeasibilityWhether you can actually staff that control at your real volume — and if not, the largest volume you can.
5. The Feasibility TestWhether the role is commissionable as scoped, and the four moves available when it is not.
6. The Value LedgerThe economics with the terms that normally go missing: the counterfactual, whether released capacity converts to anything, and the cost of the errors that slip past review.
7. Authority and PortfolioSeven distinct action powers with thresholds and expiry dates — then which role to commission first, given that senior review time is a single shared pool every AI role draws from.