AI Amplifies Additive Bias -- Subtraction Offers Competitive Advantage

The seductive ease of creation in the AI era has amplified our innate tendency to add, not subtract, leading organizations and individuals toward bloat and complexity. This conversation with Leidy Klotz, author of Subtract, reveals the hidden consequences of this additive bias, particularly as generative AI lowers the barrier to producing more content and code. The core thesis is that true elegance and effectiveness lie not in accumulating features or tasks, but in the disciplined removal of what is no longer necessary. This insight is critical for leaders and practitioners navigating the AI landscape, offering a strategic advantage to those who can resist the urge to simply "add more" and instead cultivate the difficult but rewarding practice of subtraction. Understanding this bias is the first step toward designing systems and environments that protect focus and true value.

The Hidden Cost of "More"

The pervasive bias toward addition, a fundamental human tendency, is amplified in the age of generative AI, leading to organizational bloat and a neglect of essential maintenance. Leidy Klotz, author of Subtract, argues that our default problem-solving approach is to add, a pattern he observed even in a simple Lego experiment with his son. While adding feels productive, it often obscures the more elegant, though sometimes more difficult, path of subtraction. This tendency is particularly dangerous in organizations, where the accumulation of processes, meetings, and tools can lead to "indigestion" rather than "starvation," as Bob Sutton notes. The ease with which AI can generate code, content, and workflows exacerbates this issue, making it easier than ever to add complexity without considering its downstream effects.

"More organizations die from indigestion than starvation. Over time, organizations add and add and add and spend less time thinking about subtraction, and that gets them into the situation where they're bloated."

-- Leidy Klotz

The implications for organizational design are profound. Instead of continuously adding new features or processes, leaders must actively design systems that encourage subtraction. This means questioning the necessity of existing elements and creating mechanisms to remove them. The Scandinavian design principle, "Things aren't done when you can't add more to it, but when you can't take more away," offers a stark contrast to the prevailing additive culture. Achieving this subtractive elegance requires significant upfront effort--solving the problem and then refining it by removing unnecessary components--a process often more demanding than simply adding more. This is where a competitive advantage can be built; by embracing the difficulty of subtraction, organizations can achieve a leaner, more effective state that others, caught in the cycle of addition, will struggle to replicate.

The AI Paradox: Amplifying Bias, Enabling Solutions

Generative AI presents a double-edged sword: it dramatically lowers the cost of production, thereby amplifying our inherent additive bias, yet it also offers powerful tools to counteract this very bias. As Klotz explains, AI can be used to exacerbate our tendency to add, making it easier to generate more content, code, and features with less effort. This abundance, however, comes at a cost. The sheer volume of information and options can overwhelm, making consumption more difficult than production. This flip in cost dynamic--where consuming information becomes more expensive than producing it--is a critical consideration in the AI era.

However, AI can also be a powerful ally in practicing subtraction. Klotz suggests using AI for "subtractive prompts," such as asking it to condense text or clean up technical debt. This reframing is crucial: instead of using AI to simply create more, we can leverage it to create less but better. The challenge lies in curation. As one speaker notes, simply uploading more information into an AI model doesn't necessarily improve results; it's about providing the right information. For instance, when using AI for marketing copy related to a book, providing a relevant chapter is more effective than uploading the entire book. This highlights a nuanced understanding of context engineering: avoiding both excessive information that distracts the AI and insufficient information that yields poor results. The goal is to guide the AI with precision, recognizing that its output is a reflection of the input and the prompt’s design.

"As far as, I think, obviously, AI is a tool, right? And it can be used to exacerbate these basic tendencies that we have, and it can be used to correct them."

-- Leidy Klotz

This leads to a critical insight: the "about me" or system prompt for AI becomes a form of "Odysseus contract." Just as Odysseus had himself tied to the mast to resist the sirens' song, we must instruct AI to counteract our own cognitive biases, such as the additive bias or satisficing. This requires a deliberate act of self-awareness and architectural design for our interactions with AI. By codifying our principles and values into these prompts, we can use AI not just as a tool for execution, but as a mirror to reflect and reinforce our desired behaviors and decision-making frameworks. This proactive approach is essential for maintaining agency in an increasingly AI-driven world.

Architecting for Subtraction: Guardrails in an Abundant World

The difficulty of practicing subtraction stems from both cognitive and cultural factors. Cognitively, we are less likely to consider subtraction as a solution. Culturally, additive behaviors are often more visible and socially accepted, leading to a bias towards "doing more." To counteract this, Klotz advocates for building "cues" into everyday processes, rather than relying on spontaneous moments of subtractive insight. This might involve creating "stop-doing" lists alongside to-do lists, or building automatic expiration dates for projects and meetings. The "default dead" principle, where projects are assumed to be killed unless actively chosen to continue, is a powerful organizational mechanism for enforcing subtraction.

The proliferation of AI tools presents a new frontier for this challenge. The ease with which AI can generate output means that the temptation to add is greater than ever. To navigate this, individuals and organizations must architect their environments to protect against these impulses. This involves establishing "big rocks"--core priorities--and ensuring they are addressed before the "sand" of less important tasks or AI-generated content fills up available time and cognitive space. This is the essence of "big rocks before bots." For example, setting strict boundaries around personal time, like a non-negotiable date night, acts as a crucial guardrail against the seductive allure of constant productivity enabled by AI.

"The interventions that you would make for your Odysseus tied to the mast could deal with multiple biases, right? Yes. So it's not like you need to have an intervention for each one. You could just think about like, 'Okay, here's, here are the default patterns that are going to be the ones that most get in my way.'"

-- Leidy Klotz

Furthermore, the concept of "space before screen" encourages a deliberate re-engagement with the physical world, counteracting the passive consumption often facilitated by screens. This isn't about rejecting technology, but about designing intentional interactions. Using AI to remind us to pause, to check in with our surroundings, or to question the default output, can help maintain a balanced relationship with these powerful tools. Ultimately, the practice of subtraction requires conscious design, both in our organizational structures and in our personal habits, to ensure that we are directing technology toward our goals, rather than being directed by it.

Key Action Items

• Implement "Stop-Doing" Lists: Alongside your to-do list, dedicate time weekly or monthly to identify and actively stop doing tasks, meetings, or processes that no longer serve a purpose. Immediate Action.
• Establish "Default Dead" for Projects/Initiatives: For new projects or recurring meetings, set automatic expiration dates. Require a conscious decision to renew or continue, rather than assuming perpetual existence. Implement within the next quarter.
• Leverage AI for Subtractive Prompts: Actively use AI tools to condense information, identify redundancies, or remove technical debt. Frame prompts around reduction and efficiency, not just creation. Ongoing practice.
• Define AI System Prompts as "Odysseus Contracts": Codify your core values and biases into AI system instructions to guide its behavior and protect against your own additive or satisficing tendencies. Develop and implement within the next month.
• Prioritize "Big Rocks Before Bots": Identify your most critical personal and professional priorities (your "big rocks") and schedule dedicated time for them before engaging with AI-generated tasks or other demands. Immediate Action.
• Practice "Space Before Screen": Before engaging with digital devices, consciously take a moment to observe your physical surroundings. This builds a habit of presence and counteracts passive digital consumption. Daily practice, pays off over time.
• Conduct Regular "Subtraction Audits": Periodically review your workflows, toolchains, and commitments to identify areas where complexity has crept in and can be systematically removed. Schedule quarterly audits.