Workflow Depends Too Much on People: The Design Problem Behind Every Bottleneck

2026-09-23

Executive Summary

• When a workflow “gets stuck,” the most common cause isn’t laziness or irresponsibility — it’s a process design that places too many decision points in the hands of too few people.
• McKinsey’s global survey of more than 1,200 executives found that only 37% of organizations said their decisions were both fast and high-quality; managers reported spending roughly a third of their time on decision-making, and judged more than half of that time to be ineffective.
• There are three common patterns of human dependency in workflow: single-point approval, memory-based tracking, and chase-driven progress.
• The real cost of waiting isn’t just time — it’s defensive behavior: people start asking for extra sign-off “for the record” instead of taking ownership of a decision.
• Reducing human dependency doesn’t mean removing people from the process. It means redesigning where and how many decision points actually require them.

There’s a familiar line in almost every operations meeting: “that’s waiting on [name] to approve it.” If that person is in a meeting, traveling, or simply forgot — the entire chain of work behind it stops, even when nothing genuinely complicated needs resolving.

The usual organizational response is to blame the individual: “they’re too slow to approve things,” or to send more reminders. That’s the wrong lens. When a process repeatedly stalls waiting on a person, it isn’t a work-ethic problem — it’s a systematic sign of a process design flaw, repeating itself over and over.

→ Related: From Approval Workflow to End-to-End Workflow

What a Human Bottleneck Actually Is

A human bottleneck occurs when a process is designed so its progress depends on one specific person’s availability at one specific moment — instead of depending on whether the information and conditions needed to decide are actually ready.

The key distinction: a step that requires human judgment isn’t automatically a problem. The problem appears when only one person can make that call, and there’s no mechanism for what happens when they’re absent, overloaded, or simply forget.

McKinsey’s global survey of more than 1,200 executives (published in “Decision making in the age of urgency,” 2019) produced a notable finding: only 37% of surveyed organizations said their decisions were both fast and high-quality. Managers reported spending roughly a third of their working time on decision-making — and judged more than half of that time to be ineffective. The research also found a correlation between speed and quality: organizations that decided faster tended to report better decisions, not the trade-off between speed and care that many people assume exists.

That’s evidence for an important point: slower doesn’t mean more careful, and faster doesn’t mean sloppier. The problem lives in the architecture of decision-making, not in how “careful” or “rushed” people are being.

Three Common Dependency Patterns

1. Single-point approval. A critical step in the process has exactly one person with the authority or the information needed to decide. When that person is on leave, traveling, or changes roles, the entire process stops — not because the process is missing, but because it was never designed to account for that person’s absence.

2. Memory-based tracking. No one formally “owns” tracking a process’s progress — instead, a handful of people in the organization keep it in their heads: where things stand, who needs a nudge. When that person is pulled into something else, or leaves the organization, the tracking disappears with them.

3. Chase-driven progress. The process only moves forward when someone actively messages, calls, or physically tracks someone down to prompt the next step. If nobody chases, the process doesn’t move on its own — regardless of whether enough time or the right conditions to proceed have already passed.

These three patterns typically coexist in the same process, and they reinforce each other: because things must be remembered and chased, the organization becomes more dependent on a handful of people who “know how things work” — and the more dependent it becomes, the more fragile the workflow gets whenever they’re unavailable.

The Cost of Waiting

The most visible cost is time — a request taking three days instead of three hours. But two less-discussed costs are usually larger:

The cost of defensive behavior. When decision-making is slow and unclear, people naturally protect themselves: asking for extra confirmation by email “to have it on record,” escalating decisions upward even when they have the authority to make them, or delaying action until they’re certain no one could hold them accountable. This behavior doesn’t come from a lack of skill — it’s a rational response to operating inside a system where accountability isn’t clearly assigned and decision speed can’t be relied on.

Accumulated opportunity cost. A two-day delay at one approval step rarely causes a serious immediate consequence. But when dozens of processes across the organization each have a few similar waiting points, the total “dead time” across the whole operation adds up to a significant cost — one that never shows up on any financial statement, which is exactly why it rarely gets taken seriously.

Worth noting: these costs grow with organizational size. As a company grows more layers and more waiting points, human dependency that was manageable early on becomes a genuine drag on operations — unless it gets redesigned in time.

How to Reduce Dependency Without Removing Accountability

Reducing human dependency doesn’t mean taking people out of the process — for many decisions, human judgment is genuinely necessary and shouldn’t be replaced. The problem is how people are involved. There are four directions worth pursuing:

1. Delegate by threshold, not by title. Instead of routing every request to one person simply because of their position, define which threshold — value, risk level, transaction type — genuinely needs that person’s judgment, and which can be handled by someone else or by a pre-defined rule.

2. Build formal backup, not implicit backup. If a decision genuinely can only be made by one person, that’s an operational risk that should be documented and planned for — not something the organization quietly hopes will never come up.

3. Let the system track progress instead of relying on memory. A process’s status should be recorded and visible through the system, so anyone can know “where this stands” without having to find the one person handling it.

4. Let the system prompt instead of waiting to be chased. If a step is ready to proceed but nobody has acted on it, the system should flag it automatically — this is exactly where reducing human dependency starts to overlap with moving toward event-driven workflow.

None of these four directions require complex technology to start. The first step is simply: list your most important processes, and for each one, ask “if the person handling this step were out for a week, what would happen?” If the honest answer is “everything would stop,” that’s the process to redesign first.

→ Related: Next-Generation Workflow: When AI and Organizational Knowledge Change How Work Operates

Conclusion

A workflow that depends too much on people isn’t a sign that people in the organization lack accountability — it’s a sign the process was designed to bet on the continuous availability of a few individuals. That’s a design risk, one that can be identified and reduced systematically, without waiting for it to cause a visible failure first.

This article is part of a series on workflow, AI adoption, and operational management for manufacturing SMEs.

Related articles:

• From Approval Workflow to End-to-End Workflow
• You Have Workflow. Why Is Work Still Moving Slowly?
• Next-Generation Workflow: When AI and Organizational Knowledge Change How Work Operates

→ Complete the Digitalization Readiness Assessment