Novari Edge · The edge case practice
Every automated process reaches work its rules cannot resolve. As the technology improves, that boundary moves outward, and every case beyond it lands on a person who may not be equipped, or permitted, to close it.
“…the designer who tries to eliminate the operator still leaves the operator to do the tasks which the designer cannot think how to automate.”
Lisanne Bainbridge, Ironies of Automation, 1983
86%
of organizations plan to increase AI investment.
21%
are redesigning end-to-end processes with AI at the core.
<10%
are redesigning roles, while over 40% are upskilling people.
Accenture, From early impact to enduring advantage, 2026. Survey of 3,650 executives across 20 industries, fielded January 2026.
The edge case
In an operation that keeps changing, it is a standing output. Handling it is what the people are for.
What the technology delivers
It executes what has been specified, faster and at greater volume than any team. That half is real and worth paying for. It is also the half somebody had already written down.
Why the remainder never closes
It does not consume them. The process now runs at a speed and volume the manual version never reached, and that scale produces failure modes nobody had met when the rules were written.
“Although [the new factory layout electricity made possible] was clear enough in principle, the relevant point is that its implementation on a wide scale required working out the details in the context of many kinds of new industrial facilities.”
Paul A. David, on what paced electrification four decades after the first central station opened. The Dynamo and the Computer, 1990.
Why the return stalls
Every deployment makes three decisions whether anyone owns them or not: which project is worth doing, how the work is redesigned around what the technology hands back, and who has the authority to close what it cannot. Left unowned, each one defaults. We develop the capability to make each one deliberately, in order.
01 · Executives · Days
Which processes are worth deploying against, what each one will hand back, and whether the return survives the work it creates.
The default: a pilot chosen for how well it demonstrates.
02 · Managers · Weeks
Process design around what the technology actually produces: where cases route, who decides what and at what limit, and how a resolution becomes a permanent fix.
The default: the old process with a model bolted onto it.
03 · Edge case handlers · Months
Enablement for the people already catching the cases, so a case is resolved at the desk it lands on rather than escalated up and out of the process.
The default: a redesigned process and nobody able to close a case.
Who closes the case
The coordinator who knows why this account’s orders always break, and who escalates today because deciding is above their pay grade. They hold the knowledge. What they lack is a method, a scope of authority, and the standing to resolve across desks. Five capabilities close that gap.
01
Where in the chain it broke.
02
Judgment on incomplete facts.
03
Act inside a stated scope.
04
Across desks and companies.
05
Fix, compensate, or document.
Before you book
01
02
03
04
If you can answer all four with a name and a number, you probably don’t need us. If the answers live in one person’s head, start with the audit.
Where it starts
What falls out of your processes today, what it costs you, which of it is worth engineering away and which is worth staffing properly. That measurement is the only honest basis for choosing a first project.
Scope
On a live process, it measures what the process hands back that nobody owns. Before a project is chosen, it baselines each candidate so the choice rests on a number.
What it returns
How many cases fall out, what each one costs to close, and which should be engineered away and which staffed properly. An operations measurement, not a maturity score.
What it needs
One quarter of exception or escalation logs, and a day where the cases land. It is a separate engagement and does not require a Novari subscription.
Sometimes the finding is that you need better technology. Sometimes it is that nobody in the building is permitted to make a decision. An audit that cannot return both is not worth running.
Where Novari fits
Novari Core
Novari Desk
Novari Edge
Book a 30-minute call
Tell us which process you have in mind. Thirty minutes on that process: what falls out of it, who catches it, and whether an audit is worth running.
Sources: Accenture, From early impact to enduring advantage, 2026 (3,650 executives, January 2026). Lisanne Bainbridge, Ironies of Automation, Automatica, 1983. Paul A. David, The Dynamo and the Computer, American Economic Review, 1990.