Skip to content
In this article

Ideas and terms

The AI profit and loss view

Three columns of value against the full cost

Published 9 September 20262 min read

In one paragraph

The AI profit and loss view is a one-page way of reporting what an AI system delivers, set against everything it costs. It sorts the value into three columns: what it earned, what it saved, and what it helped avoid, and nets that against the full cost of running it. The result is a single, readable statement in the terms a finance team already uses, rather than a set of separate claims that are hard to compare.

Why it matters

An AI system can be genuinely useful and still be hard to justify in writing, because its benefits tend to show up scattered across different teams and different kinds of value: some of it is new revenue, some of it is time saved, and some of it is a problem that was headed off before it happened. Reported separately, these are difficult to weigh against each other or against the cost of running the system. Brought together in one place, against a single cost figure, they become a decision a finance team can actually sign.

How it works

  • Three ledgers: what the system earned, what it saved by removing effort or delay, and what it helped the organisation avoid, such as a mistake caught before it reached a customer.
  • Cost per outcome as the unit: the full cost of ownership divided by the number of cases, documents, or decisions the system actually handled.
  • A net figure: the value across the three ledgers, set against that full cost, rather than against a single line such as a licence.
  • A threshold written down in advance: the point at which the balance would tip the other way, so the decision is not made twice.
  • A verdict in three shapes: grow the system, adjust how it is used, or set it aside, with the reasoning next to the number.

What it looks like in practice

Say an assistant is helping a team draft routine correspondence. Over a measurement period, the value shows up as time saved on drafting, a small reduction in cases that had to be reopened, and a handful of errors caught before they went out. None of those on its own would justify a line in a report. Set against the full cost of running the assistant, including the people who reviewed its drafts, the three figures together give the team a single number to weigh, and a reasoned decision on whether to expand its use.

How this connects to our work

The AI profit and loss view is the format we use to report the result of AI cost and model optimisation, once a baseline has been captured and the full cost has been counted. It is written so a finance team can check every figure in it.