What it is
As agents become capable, the limiting factor stops being how much work the machines can do and becomes how much of that work a human can meaningfully attend to. A single practitioner now runs many agents at once — coding sessions, cloud workflows, review agents, chat bots, transcript processors, background monitors — and the problem is not a shortage of machine output but a flood of it arriving without an interface that fits human attention. Overload shows up in recognizable symptoms: too many finished sessions waiting for review; no clear status across many running agents; notifications firing at the wrong moments; humans repeatedly pulled into low-value decisions; growing disconnection from work running autonomously; an inability to tell which agent is stuck and needs help; token spend with no legible link to value; and finished-looking artifacts presented without evidence they actually work. The strategic response is to stop treating attention as a free, infinite input and start treating it as the organization's scarcest managed resource — budgeted, allocated deliberately to the highest-stakes judgments, and protected by interfaces that filter, summarize, and prioritize so that a human's gaze lands where it changes the outcome and nowhere else.
When to reach for it
- When a person is supervising more than one or two agents and the volume of finished work has begun to exceed what they can actually review.
- When you notice review backlogs growing, sessions finishing faster than anyone can inspect them, or "done" artifacts piling up without evidence.
- When deciding what to build next: ask whether the bottleneck is more machine capacity or better allocation of the scarce human attention you already have.
- When designing an agentic workflow: budget human attention explicitly, the way you budget compute or money, and route it toward the decisions where judgment is the real constraint.
- When token spend climbs without a clear story about the value produced, treat that as an attention-allocation failure, not just a cost problem.
When NOT to
- Do not use scarcity of attention as an excuse to disengage entirely — autonomy without any human inspection trades an attention problem for a trust-and-correctness problem.
- This is not a mandate to inspect everything; the point is selective attention on what matters, not exhaustive review of everything an agent produces.
- Do not over-build attention-management infrastructure for a single agent or a low-stakes workflow where a glance at the output is enough.
- Avoid treating attention purely as a throughput metric to maximize; the goal is judgment landing where it counts, not the most artifacts reviewed per hour.
Exemplars
- Simon, "Designing Organizations for an Information-Rich World," in Greenberger (ed.), *Computers, Communications, and the Public Interest* (Johns Hopkins Press, 1971) — https://conversableeconomist.com/2015/08/17/economics-of-information-overload-thoughts-from-herb-simon/ — origin (link is a secondary essay quoting the chapter; the 1971 original is not stably hosted): coined "a wealth of information creates a poverty of attention," the foundational statement that attention is a scarce resource requiring deliberate allocation
- Davenport & Beck, *The Attention Economy: Understanding the New Currency of Business* (2001) — https://archive.org/details/attentioneconomy00thom — classic antecedent: brought Simon's attention-scarcity frame into management practice, arguing attention is the new currency of business that organizations must consciously manage