Best at Work Insights: AI alone can't improve your company - Minnesota Lawyer

TEXT ANALYSIS: "AI alone can't improve your company"

The Dissection

This is a people-management consultant's lament dressed as data journalism. Zepeda uses Gartner's 80% headcount reduction finding as a springboard to argue for better change management, training, and human integration of AI tools. The piece reads as a genuine concern from someone who has spent 15 years studying organizational behavior and sees organizations fumbling the transition. The Klarna and IBM re-hiring anecdotes are real and useful. The framing — that the gap between AI intention and execution is a people problem — is sincere.

The Core Fallacy

The article treats the headcount-cutting pattern as a failure of execution, when it is more accurately a structural symptom of the Discontinuity Thesis playing out in real time.

Zepeda correctly diagnoses that companies are cutting headcount without capturing returns. He attributes this to poor change management, insufficient training, and CFO laziness. This is the classic "implementation gap" frame — the technology works, organizations are just too stupid or short-termist to use it properly.

But this framing misses the mechanism. The reason companies cut headcount even when returns don't improve is not primarily that they lack training programs. It is that:

1. Headcount reduction creates immediate, measurable budget room. AI tools are uncertain; laid-off salaries are certain. CFOs optimize for certainty.
2. The "returns" being measured are the wrong returns. If you measure "financial returns" as revenue per employee, you are measuring the numerator effect (same revenue, fewer people) not the denominator problem (declining consumption as mass employment erodes).
3. The companies that cut less and performed better are not proving that human capital is irreplaceable. They are proving that premature cuts create operational disruption that temporarily damages performance. This is a lag effect, not a structural rebuttal.

The article essentially argues: "If you integrate AI properly with your people, you'll get better returns." This may be true at the firm level in the near term. It is irrelevant at the systemic level. Even perfectly executed AI integration across all 350 surveyed companies produces the same result: fewer humans in productive roles, same or better financial metrics for the firm, accelerating collapse of the wage-consumption circuit.

Hidden Assumptions

1. Firm-level optimization is the right unit of analysis. The article assumes that if each company manages AI better, the system is fine. It never asks what happens when all companies succeed at this.
2. Organizational learning will solve the problem. The Klarna/IBM reversals are framed as lessons that will propagate. They will not propagate fast enough. The window between AI capability advancement and organizational learning speed is widening, not narrowing.
3. The "institutional knowledge, relationships, and trust" that remain will be sufficient. Zepeda treats these as durable assets. Under DT logic, they are temporary moats, not structural defenses. When the Sovereign class consolidates AI capital, relationships with customers, employees, and partners become negotiable, not irreplaceable.
4. The human connection to work "still matters." This is true right now. It will be less true as AI agents handle coordination, judgment, and client relationships. Zepeda's entire framework depends on humans remaining necessary for the "real conversation" that drives change. That conversation is increasingly optional.

Social Function

This article performs mid-level consultancy reassurance theater. It tells middle managers, HR leaders, and organizational development professionals that their expertise still matters in the AI era — specifically because organizations are "doing it wrong." It is professionally self-serving (Zepeda sells consulting services) and emotionally calibrated for the audience that reads Minnesota Lawyer: legal and business professionals who want to believe thoughtful human judgment remains central.

It is not disinformation. It is not malicious. It is partial truth elevated to systemic conclusion — the most seductive and dangerous epistemic failure mode.

The Verdict

The article correctly identifies that AI deployment without organizational redesign produces operational dysfunction and no financial upside. It incorrectly concludes this is the core problem. The core problem is that even successful AI integration accelerates productive labor displacement, and the firms that "get it right" in the near term are racing each other toward the same cliff.

Zepeda is advising organizations to treat AI like a high-performing new hire. The more accurate metaphor: they are learning to use a chainsaw while standing on a floor whose structural supports are being removed. The chainsaw proficiency is fine. The floor is still collapsing.