---
title: "You Can’t Improve What You Can’t See: Why Knowledge Work Could Get Worse Before It Gets Better"
description: Explore the challenges of invisible knowledge work and learn how to enhance visibility and efficiency in digital environments for better outcomes.
image: https://tech.lean.org/hubfs/You%20Can%E2%80%99t%20Improve%20What%20You%20Can%E2%80%99t%20See-1-LinkedIn.png
---

[![LEI LeanTech Logo](https://tech.lean.org/hubfs/raw_assets/public/tech-lean-org/images/lean_AI_logo_wht.png) LeanTech](https://tech.lean.org/)

[Lean Practice](https://tech.lean.org/lean-practice) [Lean AI smart\_toy](https://tech.lean.org/lean-ai) [Journal](https://tech.lean.org/journal) [lean.org arrow\_outward](https://lean.org) [Start a Conversation](https://tech.lean.org/book)

menu

[Lean Practice](https://tech.lean.org/lean-practice) [Lean AI](https://tech.lean.org/lean-ai) [Journal](https://tech.lean.org/journal) [lean.org](https://lean.org) [Start a Conversation](https://tech.lean.org/book)

mail

### Weekly Insights

Join 30,000+ practitioners. No spam.

![Background](https://tech.lean.org/hubfs/raw_assets/public/tech-lean-org/images/backgrounds/bg1.png)

[arrow\_back Back to Journal](https://tech.lean.org/journal)

# You Can’t Improve What You Can’t See: Why Knowledge Work Could Get Worse Before It Gets Better

![Steve Pereira](https://tech.lean.org/hubfs/steve-pereira.png)

Steve Pereira

Sep 30, 2026, 5:00:00 AM

![](https://tech.lean.org/hubfs/You%20Can%E2%80%99t%20Improve%20What%20You%20Can%E2%80%99t%20See-1-LinkedIn.png)

![Picture 1](https://tech.lean.org/hs-fs/hubfs/Picture%201.png?width=2475&height=1145&name=Picture%201.png)*Invisible systems of knowledge work are likely to produce more slop than value*

Lean thinking has consistently championed standardizing work, placing local efforts in a global context, and surfacing problems so the people closest to the work can be empowered to improve. These needs exist in both the world of atoms and the world of bits. As we apply lean thinking more broadly and deeply to knowledge work, making that work visible becomes a fundamental necessity.

This two-part series explores how digital environments too often create a "black box" of invisible work, fueling a cycle of firefighting and wasted capacity. Part 1 examines the core divide between visible and invisible work, identifying the "time thieves" that keep teams in a state of perpetual emergency. Beyond visibility, Part 2 moves to action, detailing how to establish the shared mental models, metrics, and response mechanisms necessary to transform hidden waste into observable, manageable flow. To begin, we need to challenge the status quo of opaque barriers surrounding digital work.

Walk onto a construction site as the foreman and find every worker standing still, glued to their phones. You would know in a second that nothing was being built. In the physical world, we can usually see when work is stalled, struggling, or blocked. An assembly line is clearer still: a stopped line means stopped progress. Walk into a software team’s office and everyone is typing, which tells you nothing about whether anything is moving. Modern knowledge work and digital transformation brought real advantages to the global economy. Visibility was rarely one of them. We have to apply visibility deliberately, using visualization, measurement, and collaborative mapping to see what the work itself no longer shows us.

### **The Fundamental Divide: Visible vs. Invisible Work**

The most significant leaps in workflow optimization occurred in physical domains. Toyota mapped material and information flow on the shop floor, and Mike Rother and John Shook later named and popularized that practice as value-stream mapping in [Learning to See](https://www.lean.org/store/book/learning-to-see/).[\[1\]](https://tech.lean.org/journal/you-cant-improve-what-you-cant-see-why-knowledge-work-could-get-worse-before-it-gets-better#_ftn1) In manufacturing, what one person notices is easy to show to everyone else. When anyone walks onto a factory floor, tangible problems are instantly recognizable: a chemical spill, an unnatural accumulation of inventory, or a machine marked with a red flag to indicate it is broken. Workers clustered in one area while another remains empty immediately signal an imbalance. These cues make it easy for a team to agree on what is actually happening, which has to happen before anyone can fix it.

In contrast, digital and knowledge work are fundamentally invisible. Nothing moves from a truck to a conveyor belt to a finished product. The work sits behind screens and abstractions. Two coworkers can sit with their laptops touching and operate in entirely different universes because the information they view and deem important does not overlap or clearly connect to a larger whole.

Even in a factory, kanban (a signal that prompts action) exists because material flow on its own was too distributed and too inconsistent to see. It turns out we’re just not good at seeing what we’re not looking for. We can’t identify what we have no frame to name. Taiichi Ohno (considered to be the author of the Toyota Production System) stood and watched production for hours before he could see the waste in front of him. Turning that insight into something other people could apply consistently took Toyota years more. Waste is invisible when it is under your nose and you have no frame for it, on a factory floor as much as in a digital workflow.

### **The Perils of Invisible Work**

When work is made of bits and has no physical constraints, a specific set of failures follows:

- *Lack of unified mental models:* Teams working from fragmented information duplicate each other, contradict each other, or create the wrong thing. Without a common map to serve as a reference point, assumptions fill in the gaps.
- *Misguided local optimization:* Speeding up one step without end-to-end visibility moves the queue downstream, where it belongs to someone else and shows up in nobody’s metrics. Without a broad view of the value stream or value-stream network, natural behavior is to improve what’s readily apparent. We fall victim to the streetlamp effect (i.e., looking where it’s easy to spot *something*, albeit not the right thing).
- *"Black Box" perception:* Stakeholders outside delivery cannot see complexity, technical debt, or real capacity, so they reasonably assume there is not much of any. In the world of knowledge work, we can’t see the mechanisms, so attention is focused on the visible contributors rather than the invisible system of work.
- *Severe cognitive load and context loss:* Every handoff loses context, and the cost comes back later as coordination overhead and exhaustion. The workload becomes juggling and coordinating an intangible mountain of complicated information, which too often seems accepted as the cost of doing digital business.
- *Ignored priorities and hidden work:* Technical debt and security have no queue of their own, so they grow like dry brush waiting to be ignited by a spark. Debt without intentional investment or a ledger to track it spirals out of control and no alarm bells ring until it’s too late.

### **The Doom Loop of Invisible Work**

The clearest symptom of invisible work is reactive firefighting. With insufficient measurement and no visible workflow, organizations fall into a vicious cycle dictated by unplanned work. Improvement work never starts because the day is already spent on crises. Many knowledge work and tech teams are not smooth systems that hit the occasional problem. They run in a more or less permanent state of emergency. When an organization spends all its energy putting out fires, it lacks the capacity to step back and analyze where the fires are starting or how they are spreading.

In Making Work Visible,[\[2\]](https://tech.lean.org/journal/you-cant-improve-what-you-cant-see-why-knowledge-work-could-get-worse-before-it-gets-better#_ftn2) Dominica DeGrandis names five “time thieves” that drain capacity, force context-switching, and burn teams out:

![Screenshot 2026-09-29 at 9.33.15 PM](https://tech.lean.org/hs-fs/hubfs/Screenshot%202026-09-29%20at%209.33.15%20PM.png?width=848&height=213&name=Screenshot%202026-09-29%20at%209.33.15%20PM.png)

 

- Too much work in progress (WIP) — work that has started but not yet finished.
- Conflicting priorities — projects and tasks that compete with each other. This gets worse when nobody will say what matters most.
- Unknown dependencies — something nobody knew had to happen before the work could finish.
- Unplanned work — interruptions that prevent you from finishing something.
- Neglected work — partially completed work that sits idle.

Put the five thieves in motion and they form a reactionary loop that feeds a growing fire. Knowledge work has been running the loop for decades, lacking the visibility to see its cost or how to escape it.

![Screenshot 2026-09-29 at 9.33.23 PM](https://tech.lean.org/hs-fs/hubfs/Screenshot%202026-09-29%20at%209.33.23%20PM.png?width=1506&height=646&name=Screenshot%202026-09-29%20at%209.33.23%20PM.png)

1. Leadership pushes against slow delivery. Contributors protest, but lack convincing data to make their case.
2. Work runs forward blindly until a hidden dependency, an interruption, or old technical debt trips it up. The block is sometimes raised. More often it is worked around with more debt.
3. Blocked work is set aside rather than finished, and new work starts on top of it. WIP goes up.
4. Coordination overhead grows with WIP, delivery slows, and priorities get reshuffled to force progress. Nothing is ever really removed or set back on the shelf.
5. The blocked work is cleared or priority shuffled. Work that is picked back up after weeks away has lost context, and under pressure to catch up, it ships rushed.
6. Rushed work comes back as bugs, rework, and more debt. These interrupt the next round, delivery slows further, and leadership steps in to push harder. Back to the start of the loop.

Every turn of the time-thieves loop raises WIP, and WIP is the thief that feeds the other four.

Today, AI is feeding the fire rather than fighting it. Generating more work in less time raises WIP, and the other four conditions all worsen as WIP climbs.

### **Invisible WIP Invites More**

Nobody would think of tasking a construction crew to add another bathroom as the paint is drying on a new house or asking a mechanic to fix two cars at once. The digital equivalents happen constantly, and nobody notices.

Two conditions uniquely plague knowledge work to make this worse:

1. Poor implementation of project, scrum, or kanban workflow provides a binary view of work in progress until it’s not. Work is rarely visible as 20% or 80% complete. Work is done when it’s done, and it could take an epiphany or just the right count of trial and error to make it over the finish line.
2. Because work is invisible, contributors are incentivized to seem busy, and disincentivized from careful thought, problem solving, and risk reduction. Fires of various sizes are seen as equal and equally urgent, leading to interruptions and heroics. Contributors with a clear view of risks and debt struggle to communicate effectively to leadership, leading to failures foreseen well in advance and “Chicken Little” or “Peter and the Wolf” dynamics.

Until we design workflow and measurement to address these conditions directly, there are too many ways these conditions can create and feed the time-thieves loop.

### **The Poor Proxies**

It’s common to see visibility tackled with approaches that don’t address the core of the problems that invisible work hides:

- A tools-first, “build a dashboard (or data mesh) and they will come” without a mechanism connecting measurement to empiricism and playbooks to guide action,
- Metrics based on “industry best practices,” like DORA (DevOps Research and Assessment), without a theory of measurement to justify their collection or use,
- A map that collects exhaustive detail yet buries the lead, and
- A clear map that doesn’t prompt action because driving factors like ownership and strategy are absent or disconnected.

Seeing effectively requires putting information in context and providing guidance on what is safe to ignore, not just what deserves attention. It’s prioritized, categorized, compared, and attributed to greater value and larger trends.

### **Why Visibility Matters More Than Ever**

AI and automation make the invisible work problem much more urgent. AI generates code, text, and data at almost no marginal cost, and does nothing for the review, approval, integration, and delivery that follow.

Speed added upstream of a bottleneck converts into inventory rather than output. In software, that inventory is unreviewed pull requests, untested features, and code that runs on one laptop. It piles up the way pallets do, except it takes no floor space, so nobody walks past it and asks what it is doing there. DevOps grappled with this scenario nearly 20 years ago, far enough in the past to be relearned by a new AI-native generation.

More recently, [DORA](https://dora.dev/guides/dora-metrics/) was one of the first efforts to drive visibility to digital delivery performance with its original four key metrics across throughput and quality. A fifth metric, deployment rework rate, was recently added to instrument this exact failure.

![Screenshot 2026-09-29 at 9.37.51 PM](https://tech.lean.org/hs-fs/hubfs/Screenshot%202026-09-29%20at%209.37.51%20PM.png?width=735&height=253&name=Screenshot%202026-09-29%20at%209.37.51%20PM.png)

*Without models, maps, and metrics in a connected context we work blindly.*

Accelerating output without a visible workflow floods hidden bottlenecks and buries teams in unverified work. Engineers can now "vibe code" a solution, prompting their way to software they do not fully understand and producing more waste faster than any earlier tool allowed. Making work visible is no longer just a best practice — it is the prerequisite for safely adopting AI at scale.

Once we can truly “see” and measure flow across the value stream, and across the broader network, people at every level can act in service of the whole system instead of their own immediate surroundings. That holds whether we are watching work items move across a kanban board or reading the “digital exhaust” those movements leave behind in our metrics.

In Part 2, I’ll share the most useful mechanisms I’ve found for building visibility, but also the challenges that sabotage them over and over again.

*Article and graphics developed with the assistance of AI.*

---

[\[1\]](https://tech.lean.org/journal/you-cant-improve-what-you-cant-see-why-knowledge-work-could-get-worse-before-it-gets-better#_ftnref1) Mike Rother and John Shook, Learning to See (Lean Enterprise Institute, 1999).

[\[2\]](https://tech.lean.org/journal/you-cant-improve-what-you-cant-see-why-knowledge-work-could-get-worse-before-it-gets-better#_ftnref2) Dominica DeGrandis, Making Work Visible (IT Revolution Press, 2017).

mail

### Get Weekly Insights

Join 30,000+ practitioners getting the Journal every week. No spam.

## About the Author

![Steve Pereira](https://tech.lean.org/hubfs/raw_assets/public/tech-lean-org/images/team/steve-pereira.png)

### Steve Pereira

Flow Engineering

 Steve Pereira has spent his career on a problem that lean practitioners know well but rarely solve cleanly in technology environments: you can't improve what you can't see. Working with technology and service organizations, he developed Flow Engineering, a methodology that adapts value stream mapping for software delivery and knowledge work -- where value streams are invisible and work flows through systems rather than factory floors. He co-authored Flow Engineering: From Value Stream Mapping to Effective Action and serves as a board advisor to the Value Stream Management Consortium. He leads flow engineering for LEI's LeanTech initiative, helping organizations move from mapping to action in environments where speed and complexity make standing still costly.

Share this article:

![LEI LeanTech Logo](https://tech.lean.org/hubfs/raw_assets/public/tech-lean-org/images/lean_AI_logo_wht.png)

 LeanTech

 LeanTech is a practice of the Lean Enterprise Institute, helping technology organizations apply lean thinking principles to drive sustainable transformation. 

[Visit lean.org arrow\_outward](https://lean.org)

#### Explore

[Lean Practice](https://tech.lean.org/lean-practice) [Lean AI](https://tech.lean.org/lean-ai) [Journal](https://tech.lean.org/journal)

#### Connect

[Book a Conversation](https://tech.lean.org/book) [Contact Us](https://tech.lean.org/contact) [Newsletter](https://lean.org/newsletter)

 © 2026 Lean Enterprise Institute, Inc. All rights reserved.

[Privacy Policy](https://tech.lean.org/privacy) [Contact](https://tech.lean.org/contact)

```json
{
  "@context" : "https://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "Steve Pereira",
    "url" : "https://tech.lean.org/journal/author/steve-pereira"
  },
  "dateModified" : "2026-09-30T09:00:00.124Z",
  "datePublished" : "2026-09-30T09:00:00.000Z",
  "headline" : "You Can’t Improve What You Can’t See: Why Knowledge Work Could Get Worse Before It Gets Better",
  "image" : [ "https://tech.lean.org/hubfs/You%20Can%E2%80%99t%20Improve%20What%20You%20Can%E2%80%99t%20See-1-LinkedIn.png" ],
  "mainEntityOfPage" : {
    "@id" : "https://tech.lean.org/journal/you-cant-improve-what-you-cant-see-why-knowledge-work-could-get-worse-before-it-gets-better",
    "@type" : "WebPage"
  },
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "url" : "https://tech.lean.org/hubfs/LEI_logo2+url_blue.png"
    },
    "name" : "Lean Enterprise Institute"
  }
}
```