AI’s Workslop Problem — And the Guardrails We Critically Need

The Promise — and the Reality 

Generative AI has inspired an investment boom perhaps not seen since the dot-com era. Trillions of dollars in market capitalization now ride on promises of productivity, automation, and business and lifestyle transformation. Corporate boards and C-suites are pressing for “AI-first” strategies and perhaps even penalized by the market if they don’t show signs of keeping up.

But the reality looks far less impressive. Research at MIT suggests that 95% of AI pilots fail to scale into production. Much of what leaders are calling “AI deployment” is in fact what Harvard Business Review recently labeled workslop: machine-generated output that looks efficient but undermines quality, distracts employees, and erodes attention.

We have been here before. New technologies almost always overshoot in their early phases, generating enthusiasm out of proportion with the actual gains. The question is whether this moment becomes just another dot-com-style bubble—or something more durable.

Where AI Is Actually Working 

There are exceptions. A new HFS Research study highlights the “15% Club” of organizations achieving real, measurable returns from AI. What sets them apart?

The study suggests that these organizations are not by chasing general-purpose AI solutions but by putting in place clear leadership accountability, embedding AI into broader transformation efforts, and moving investment decisions closer to the business lines where outcomes are realized. Their approach is reinforced by flexible funding models that adapt as results emerge and by a pragmatic focus on outcome-based milestones and use cases. Perhaps what we see in this work is an elegant illustration of the success of domain-specific AI.

The lesson might be straightforward. AI without guardrails produces a lot of noise. AI with boundaries that consider the messy complexities of human-led organizations, limitations of our ability to work with automation, and a specific design focus can produce value.

A Cautionary Tale from the Road 

This lesson is not confined to business. In transportation, my MIT research has examined how automation changes human behavior. The early rollout of Tesla’s Autopilot, for example, demonstrates both promise and peril. The system could make highway driving less taxing—but it also makes drivers less attentive. We’ve even documented drivers turning around in their seats or tying a rope while Autopilot handled the car, and heard tales of  “hot-swaps” with drivers and passengers switching on the go.

By contrast, GM’s Super Cruise from the start integrated driver monitoring and support systems to help keep humans engaged. The difference is not just technical—it is policy by design. Super Cruise reflects an intentional choice to support the driver rather than sideline them.

Tesla has moved in this direction, but rather than designing around the limitations of human capabilities from the start, they took a technology-first focus that left drivers at potentially greater risk.

The business of AI faces some of the challenges that Tesla faced with Autopilot a decade ago. Letting technology vs human-focused use cases lead. Left unconstrained, this technology-first mindset risks eroding the very productivity it promises. Designed with guardrails that consider human capabilities, limitations, and values, it can amplify human performance.

From Bubble to Balloon 

Given all of this news, too many leaders, policymakers, and consumers are asking whether AI is a bubble. A better metaphor is a balloon. A bubble bursts and disappears. A balloon inflates, deflates, and rises again.

The dot-com boom of the late 1990s was a true bubble—valuations soared without business models, and when it burst, much of the capital and companies simply vanished.

Electricity, the personal computer, and the smartphone all followed balloon-like trajectories. Their value surfaced not in the initial hype cycle, but through the slow layering of infrastructure, standards, and governance that embedded them into daily life.

AI will follow the same pattern. Its lasting value will appear not through unbounded pilots, but through domain-specific deployments where attention is managed, data is disciplined, and human expertise is amplified.

The Policy and Culture Imperative 

Technology design is only half the story. Policy also shapes whether AI becomes workslop or a productivity amplifier. In driving, regulators have allowed companies to market “autopilot” systems without requiring robust driver monitoring—a choice that has contributed to public confusion, misuse, and tragedy. The lesson is clear: without policy guardrails, commercial incentives will push technology faster than society can safely absorb it.

The same applies to AI in business. Transparency around data provenance, accountability for model outputs, and clarity on human oversight are not “nice-to-haves.” They are prerequisites for trust. Regulators, industry consortia, and corporate boards need to establish standards that ensure AI supports rather than supplants human judgment.

But policy alone is insufficient. Work culture and personal responsibility matter just as much. Employees need training not only in how to use AI tools, but in when not to use them. Leaders must set norms for quality and accountability, ensuring that AI augments rather than replaces human diligence. Guardrails are not just technical or regulatory; they are cultural.

Flying Higher 

AI is not destined to fail. But neither will it succeed simply because we throw more capital at it. Business leaders who treat AI as a general-purpose magic wand are likely to waste money and time. Those who define domain-specific boundaries, implement attention-shaping guardrails, and build cultures of responsibility will be the ones to fly their balloons higher and longer.

The challenge for leaders is not whether to invest in AI. It is whether they are willing to structure it in ways that make us better. Leveraging AI to amplify human capabilities. The organizations that have the right minshift will fly the highest; those that don’t will see their balloons deflate.

Like any other line of business, weak leaders will ignore the state of the ballone, move smoke and mirrors to hide as their organizations falter. Strong leaders, however, will identify deflating balloons quickly and make the changes needed to ensure their organizations fly higher and higher.

Written by Bryan Reimer, PhD, in partnership with Magnus Lindkvist.