The Asymmetric Cost Problem: How AI-Generated Communication Is Quietly Degrading Your Engineering Organisation

Every productivity tool that reduces effort for the sender eventually increases effort for everyone else. AI-generated communication is following exactly that pattern inside engineering organisations right now. The problem is not that engineers are using AI to write messages, documents, or tickets. The problem is that the cost of producing communication has collapsed while the cost of consuming it has not, and that asymmetry is accumulating silently across every team that has not yet named it.

The Economics of Attention Are Not Symmetric

Writing a detailed Slack message, a well-structured ticket, or a considered design proposal takes effort. That effort has historically acted as a natural filter. Senders calibrate how much they ask of readers because producing the ask costs something.

AI removes that friction for the sender without removing the corresponding burden for the reader. A two-paragraph update that would have taken ten minutes to draft now takes thirty seconds. The reader still needs the same two minutes to parse it, and they still need to decide whether it warrants a response.

At the individual level, this feels like a minor inconvenience. At the organisational level, when dozens of engineers are generating more communication with less friction, the aggregate reading burden on the team rises faster than any individual sender notices.

What Workslop Actually Costs Engineering Teams

We use the term "workslop" deliberately. It describes communication that has the surface appearance of substance but lacks the density of thought that genuine writing requires. AI-generated workslop is not malicious. It is the predictable output of a tool optimised for fluency rather than precision.

The cost shows up in several places. Engineers spend more time reading to extract less signal. Code review comments become verbose without becoming more useful. Incident postmortems expand in length while shrinking in diagnostic value. Sprint retrospectives fill with well-formatted observations that nobody acted on because nobody had to think hard enough to write them to internalise the point.

The deeper cost is to the quality of technical reasoning itself. Writing forces the author to resolve ambiguity before externalising a thought. When AI resolves that ambiguity on the author's behalf, the underlying thinking may never have happened. The document exists; the decision clarity it was supposed to represent does not.

Where the Denial-of-Service Dynamic Emerges

The phrase "denial of service" is borrowed from network security deliberately. A DoS attack does not require malicious intent at the individual packet level. It requires only that the volume of incoming requests exceeds the capacity of the system to process them usefully.

Engineering teams have a finite capacity for communication processing. That capacity is already under pressure from meeting load, context switching, and the ordinary cognitive demands of technical work. When AI-generated communication inflates message volume without inflating informational value, the system begins to degrade. Engineers start skimming. Important updates get missed. Decisions that needed a thoughtful response get a quick acknowledgement instead.

The engineers most likely to keep reading carefully are also the engineers most likely to be in high-demand roles. Senior engineers and tech leads absorb the most communication tax because their judgment is most frequently sought. The people you can least afford to distract are the ones most exposed to the asymmetric cost.

Governance Responses That Actually Work

Individual coping strategies, such as inbox rules, notification muting, or personal norms around response time, do not address the structural problem. They redistribute the burden rather than reducing it. Engineering leaders need explicit governance at the team and organisation level.

The starting point is defining communication standards by context, not by individual preference. A well-formed incident report has a defined structure. A ticket that is ready for engineering has specific required fields. A design document that is ready for review has a stated decision, a set of considered alternatives, and a clear ask. These standards exist in most mature engineering organisations already. The question is whether they are enforced with the same discipline now that AI makes it trivially easy to produce something that looks compliant without being compliant.

The second lever is accountability for communication quality, not just communication volume. If your engineering culture rewards thoroughness as demonstrated by length, AI will produce length. If it rewards clarity as demonstrated by decisions made and action taken, the incentive structure works against workslop even when AI is doing the drafting.

The Cultural Decision Technical Leaders Need to Make

There is a version of this problem that resolves itself badly. Teams normalise low-density communication, senior engineers quietly disengage from written channels because the signal-to-noise ratio no longer justifies the attention cost, and important decisions migrate to informal conversations that leave no record. That outcome is not hypothetical. It is the direction of travel for teams that treat AI-generated communication as a productivity metric rather than a quality risk.

The cultural decision is whether communication is treated as a product of thinking or as a substitute for it. Engineering leaders who are explicit about that distinction, in onboarding, in code review norms, in how they respond to workslop when they see it, create the conditions for AI to be a genuine productivity tool rather than an attention tax.

This does not mean prohibiting AI assistance in written communication. It means holding the standard for what communication is supposed to accomplish and making it clear that AI is responsible for the drafting, not for the thinking that should precede it.