The Dangerous Third: What If the Work You Want to Automate Is Holding the Organisation Together?

The second piece in this series introduced the most immediately usable instrument in our FusionWork™ research: the Thirds Model. A typical week splits into roughly equal parts. Right work, the strategic and value-creating third. Wrong work, the duplication, rework and approval-chasing that consumes capacity without creating value. And invisible work, the coaching, workarounds, knowledge sharing and stakeholder management that appears in no system and no job description.

“The Thirds Model: Why AI Automates the Wrong Third of Your Workforce’s Time”.

The model’s central claim is that role-based AI (Artificial Intelligence) deployment sees only what is written down, and is therefore blind to the third that often matters most. As diagnosis, I think that claim is right, and the Workforce X-Ray exists precisely to correct it.

My interest starts where the diagnosis ends. Because once the X-Ray has done its job and the invisible third is lying on the table in front of you, a harder question appears, and the model, deliberately, does not answer it.

What is this work for?

Read the invisible list again: informal coaching, workarounds, knowledge sharing, stakeholder management, change navigation.

Four of those are capabilities. One is a symptom.

Coaching, knowledge transfer and stakeholder repair are tacit value: judgement and relationship work the formal organisation never learned to see. But a workaround is different in kind. A workaround exists because something in the formal organisation is broken. It is capacity consumed compensating for a defect, which is the model’s own definition of wrong work. It simply happens to be undocumented.

That is the seam I want to press on, respectfully, because it changes what you do on Monday. The Thirds Model classifies work by visibility. Automation decisions require classifying work by structural function. And those two axes are orthogonal.

Cross them and you get 4 categories, not 3. Visible value and visible waste, which the model already separates well. And then two kinds of invisible work that look identical in any survey and demand opposite treatment: invisible value, the organisation’s tacit intelligence, and invisible debt, the accumulated compensations for systems, processes and interfaces that don’t work. Call the second category what it is: organisational debt. Manual reconciliation between systems that don’t talk to each other. Shadow spreadsheets that exist because the official one can’t be trusted. Unofficial escalation paths that route around a broken approval chain. Humans acting as middleware between dysfunctional interfaces.

Here is why the distinction is not academic.

Take a familiar case. An analyst spends 3 hours every week manually reconciling two systems that don’t communicate. The X-Ray surfaces it. It is invisible, repetitive, and joyless, so it goes straight onto the automation list. An agent is deployed. The 3 hours are recovered. The dashboard improves.

And the organisation has just made its dysfunction permanent. The broken interface still exists. The workaround has not been eliminated; it has been made cheap, silent, and infrastructural. Nobody will ever again feel the friction that would have justified fixing the integration, because the friction is now borne by software that doesn’t complain. Worse, the analyst knew when the numbers looked wrong. The script does not. The exception-detection that was quietly bundled into the drudgery has been deleted along with it.

That is the concept I think this research programme needs to hold onto: automation can convert organisational debt into automated organisational debt. Compounding, and harder to see than it was before, because now not even a human touches it.

The opposite failure is just as available. Automate away what looks like low-value coordination and you may discover it was the tacit coaching and stakeholder repair holding a function together, exactly the breakage Cheney warned about when the person carrying invisible work leaves. Two failure modes. Opposite remedies. Identical appearance in the data.

The same logic reframes a person every organisation recognises: the one everybody calls indispensable. They know whom to call, which rule bends, which spreadsheet is actually correct, how to get Finance and Operations to agree, which system output cannot be trusted. Talent management files that person under high performer, and often they are one. A systems reading asks a less flattering question: how much of this indispensability is excellence, and how much is the enterprise requiring a human being to continuously compensate for its architecture? Often the answer is both, entangled. Which means capturing their tacit knowledge, the instinctive AI-era move, will either distribute genuine capability or encode institutional dysfunction into a system, and the capture process itself doesn’t tell the difference. The structural question does.

The sorting test starts with a single question. For every piece of invisible work the X-Ray surfaces, ask why it is invisible.

If it is invisible because it is tacit, because it lives in judgement, relationships and experience that formal systems cannot hold, you have found capability. What follows is a preservation problem: the value sits in the work itself, and automation close to it takes the value out with the effort.

If it is invisible because it is compensatory, because a formal system failed and a human quietly filled the gap, you have found debt. What follows is a redesign problem: automated in place, the compensation becomes permanent infrastructure; removed at source, the work disappears together with the defect that created it.

And expect to find both in the same piece of work. A customer escalation workaround exists because the CRM (Customer Relationship Management) workflow is broken; that is debt. But the person operating it has developed real judgement about which customers to escalate, and when; that is capability. The workaround and the wisdom arrived together, and they have to be separated deliberately, because no single automation decision treats them both correctly.

Which is why the 4 verdicts available for any piece of surfaced work, preserve, redesign, eliminate, automate, are combinable rather than mutually exclusive. The escalation workaround admits all 4 at once: preservation is what the judgement needs, redesign is what the broken workflow needs, elimination is what the compensating work gets once the workflow is fixed, and automation fits whatever routine transaction remains. Visibility alone selects none of them.

Cheney’s discipline stands: don’t automate what you cannot see, and start with one function, not the whole organisation. The X-Ray is the right instrument, and this piece is an argument for using its output more suspiciously, not for using it less.

But making invisible work visible is necessary and not sufficient. Before automating anything the X-Ray reveals, you need to know why it became invisible in the first place, because the answer determines whether you are about to scale the organisation’s intelligence or institutionalise its dysfunction.

And notice what kind of question that is. Whether a workaround is preserved, redesigned, eliminated or automated is not an analytical output. It is a judgement about the organisation’s architecture, with consequences for capability, cost and risk that outlast any deployment. Somebody has to be authorised to make it. Who that somebody should be is a question this series will keep returning to.

Print Article for this blog Download here