Software companies know how to measure labor saved.
Time per task. Transactions per employee. Cost per ticket. Orders per operator. Reports produced per week.
These measures reveal execution efficiency. They do not tell us whether the process still occupies human attention.
A workflow can become ten times faster while still requiring someone to remember it, initiate it, inspect it, and remain responsible for failure.
If the goal is absorption rather than automation, we need different measures.
Absorption transfers predictable work beneath ordinary conscious supervision. Its value appears partly in what no longer happens.
Start with required visits.
How often must someone open a system for the process to remain trustworthy? Include scheduled review, habitual checking, approval queues, status reports, and meetings whose main purpose is to reconstruct system state.
Making each visit shorter improves automation.
Making the visit unnecessary returns attention.
Next, measure interruptions.
How often does the system recruit a person? More importantly, what proportion of those interruptions did users agree deserved their attention?
Call this exception precision.
A system with low exception precision creates alert fatigue. A system that stays quiet while consequential failures remain hidden creates false silence. Both prevent absorption: one captures attention constantly, while the other teaches users to rebuild the dashboard through checking.
Measure repeated exceptions too. If the same predictable correction reaches a person every week, the product is using human attention instead of improving its model, authority, or recovery.
Then measure context reconstruction.
An alert may take five seconds to open and twenty minutes to understand. The user locates the original event, reads a thread, compares records, determines what the system tried, and identifies the actual decision.
A high-quality exception arrives with the change, consequence, evidence, attempted recovery, remaining options, and deadline. It recruits judgment without making the person become the operator again.
Another useful measure is recovery autonomy.
What proportion of predictable failures can the system diagnose, contain, and resolve inside legitimate boundaries? Do not count actions taken without a human. Count failures resolved without unnecessary attention and without exceeding authority.
Finally, measure responsibility duration.
How long does the process remain an open loop in someone's mind? Do users create side reminders? Do they return without an alert? Do they maintain parallel spreadsheets? Do they ask for confirmation that a completed process is actually complete?
These behaviors reveal attention that time studies miss.
The deepest metric is attention per outcome.
For a successful order, how many visits, interruptions, approvals, minutes of reconstruction, and unresolved reminders did people supply? For a completed project, how much conscious supervision kept commitments current? For a customer relationship, how often did someone rebuild context the system could have maintained?
The measure will never be perfectly precise. It does not need to be. Its purpose is to make attention visible as a design variable.
This can conflict with traditional engagement metrics. A valuable absorbing product may produce fewer sessions, fewer clicks, fewer notifications, and less time in the application as it improves.
If revenue depends on captured attention, the product has an incentive to preserve reasons to return. If revenue depends on maintained outcomes, resolved risk, or obligations carried, the product can benefit from disappearing.
The question is no longer only how much time the tool saves while someone uses it.
It is how much of the person's life no longer needs to contain the process at all.