Shieldbase
May 28, 2024
In an almost sci-fi scenario, organizations face disruption from AI unless they can persuade their "secret cyborgs"—employees who discreetly leverage AI—to come forward. This description accurately captures the current challenge for businesses.
The Promise and Pitfalls of Large Language Models
To understand the significance, we must recognize that while Large Language Models (LLMs) are revolutionary for individual productivity, they have yet to achieve the same impact on organizations. Early studies indicate AI can significantly boost individual productivity, with time savings ranging from 20% to 70% for various tasks and higher quality outcomes compared to non-AI methods. However, AI's inconsistency and propensity for errors limit its effectiveness as organizational software. Currently, AI excels as a personal productivity tool when used by experts in their fields.
Innovation in Secrecy
Today, billions have access to LLMs and their productivity benefits. Research across diverse professions shows that given general-purpose tools, individuals devise innovative uses that can transform their jobs. Employees are finding new ways to streamline tasks, code, and automate tedious parts of their work. Yet, these innovations often remain hidden from their companies. These employees are the "secret cyborgs," machine-augmented workers who prefer to keep their AI use under wraps.
The Secret Cyborgs
There are several reasons these cyborgs remain undisclosed, primarily to avoid trouble. Organizational policies are a significant barrier. Many companies have banned tools like ChatGPT due to vague legal concerns and regulatory uncertainties. While legal teams are cautious, AI companies are striving to make their tools compliant with laws, such as Anthropic’s HIPAA-compliant Claude AI and OpenAI’s security-focused offerings. Despite justified concerns, broad AI bans are likely temporary. Companies should consider targeted policies instead of blanket prohibitions.
Shadow IT: A Growing Concern
These bans have led employees to use personal devices to access AI at work, creating a form of Shadow IT. This hidden usage violates company policies, making employees reluctant to reveal their AI activities. Another reason for secrecy is the perception of AI-generated content. AI’s ability to produce human-like writing is powerful but loses its impact if known to be AI-generated. Research indicates that people judge AI-created content more harshly than human-created content. Unsurprisingly, many AI users, as seen in an informal Twitter poll, use the technology discreetly.
Fear of Replacement
Employees also fear that by revealing their AI usage, they might be training their replacements. If someone automates 90% of their job, disclosing this might lead to workforce reductions. Thus, it is safer for employees to keep their AI use confidential.
Unleashing the Cyborgs
Traditional organizational approaches to new technology adoption are too centralized and slow for AI. IT departments cannot easily develop in-house AI models that compete with major LLMs. Consultants and integrators lack the specialized knowledge needed to make AI work effectively for specific companies. Internal innovation groups and strategy councils can set policies but are not equipped to implement AI in daily work—only employees, experts in their jobs, can do that.
Embracing Decentralized Innovation
To harness AI's full potential, organizations need their cyborgs' help and must encourage broader AI use among employees. This requires a significant shift in organizational operations.
Identifying Hidden Talent
First, companies must acknowledge that employees with AI skills could be at any level, irrespective of their previous performance records. AI proficiency wasn’t a hiring criterion, so expertise could be widespread. Initial evidence suggests lower-skilled workers benefit most from AI, but this is still evolving. Companies should integrate as many employees as possible into their AI initiatives, offering broad training and tools like crowd-sourced prompt libraries for knowledge sharing.
Reducing Fear and Building Trust
Second, leaders must reduce the fear associated with revealing AI use. They can assure employees that AI won't lead to layoffs and promise that freed-up time can be used for more engaging projects or even early work finishes. Studies show that AI can make work more enjoyable by eliminating monotonous tasks, which can be a strong incentive. Trust and a positive culture are crucial; employees must believe that their organization values them to feel safe disclosing their AI use.
Incentivizing Innovation
Third, organizations should incentivize cyborgs to come forward and expand AI use. Beyond permitting AI use, they should offer substantial rewards for identifying significant AI opportunities. Incentives might include cash prizes, promotions, desirable office spaces, or permanent remote work options. The potential productivity gains from LLMs justify these rewards, underscoring the organization's commitment to innovation.
Addressing the Future of Work
Finally, companies need to address fundamental questions about managing productivity gains, reorganizing work, and handling AI-related risks. This includes reassessing processes made redundant by AI and managing AI-driven hallucinations and IP issues. Procrastinating on these issues will lead to worse outcomes. Embracing the "cyborgs" and collaborating with them can help build a new, more efficient organization in the AI era.
