The work elsewhere
What this platform can actually see
Across all fourteen functions the platform observes 77.9% of the work, reconstructs 14.3% by rule or cohort, and knows that 7.8% happens where it cannot read it. That last figure is 15,628 items a week. It is published here, next to the numbers it affects, rather than on a methodology page nobody opens.
Every function, and how much of it the platform reads
Ordered as the estate is ordered, not by score. Research & Development is the least observed at 30.0%, AI GBS & Global Business Services the most at 90.0%.
Every outcome chain, and how its links were joined
Here observed means a stage link joined on a shared key. Inferred means joined by rule or by cohort. The unit is a chain link, not a work item, and the two counts are not comparable.
How each figure was produced
A method is the thing to argue with. Every statement on this page names one, and the weaker methods are named as clearly as the stronger ones.
Actions
What is waiting on a person
A coverage statement is only as good as the method behind it. These are the statements whose method is the weakest, in the order a function head should challenge them.
statements carrying an uncertainty band of a quarter or more
The widest is Hire to productive at plus or minus 35%. A band this wide means the figure should be read as a range, and the product prints the band next to the number rather than behind a footnote.
functions whose unseen share came from an interview, not a count
Legal · Human Resources · Administration. Nobody counted the unseen work in these functions. Somebody was asked, and the answer is on the record with their name against it.
outcome chains with no keyed link at all
Invoice to payment · Deduction to settlement · Hire to productive. Every chain on these services is reconstructed. The coverage statement reports observed as zero rather than reporting the inferred count and calling it coverage.
function estimate nobody has put their name to
Human Resources. An attestation would not make the figure measured. It would put a person behind the estimate, which is the difference between a number and a claim.
Operations
What this desk is allowed to start
A surface that only reports is not operable. This is the work this page can set in motion, and the bound it runs into.
Trigger and bound
This surface publishes a coverage statement and lets anybody dispute the method behind it. It cannot widen coverage. Coverage widens when a source system is connected, when off-platform work is filed through intake, or when a shadow process is confirmed and counted - three mechanisms that live on other pages and write back here.
Live observability
What the record shows right now
Work items a week across all fourteen functions, split 3 ways by how the platform knows about them. The three add to 200,240, which is the estate figure carried everywhere else in the product.
Current distribution
200,240 items a week
Is policy and strategy coming to fruition
Whether the written intent is holding here
Yes on saying it. The gap is published beside the figures it affects, and it is not small.
Holding on the record
The written intent is that the platform says what it cannot see rather than reporting a clean estate. That now holds: every function and every outcome carries a statement, 15,628 items a week are declared unseen, and the mean uncertainty band across all 22 statements is plus or minus 18.0%. What has not happened is any narrowing. Every figure here is the first measurement, so there is no trend to read yet and the product does not pretend otherwise.
How to disagree with a number on this page
Coverage can only ever be estimated from the outside. Observed is a count the platform can produce from records it holds. Inferred is a count it reconstructs by rule or by cohort. Unseen is an estimate of work the platform knows exists and cannot read, and every statement carries the method that produced it and a stated uncertainty. Nothing here is rounded up.
A coverage statement is disputable by design. The way to disagree with one is to inspect the method and the basis for the unseen figure, both of which are on the statement, rather than to argue about the number.
3 statements rest on an interview rather than a count. Those are the ones to challenge first, and the function head named on each one is the person who gave the figure. 3 statements carry an attestation, and every one of them is signed by an office rather than by an individual, which is the weaker of the two.