The human factor in generative AI and the automation paradox

Effective leaders understand the automation paradox: the more sophisticated and complex technology gets, the more vital human users become to its operation. While artificial Intelligence (AI) is not new, advancement of the technology through generative AI has been so rapid that many business leaders have had trouble keeping pace with the changes.

With thanks to Jackson Bremen, Katie Mumford, and members of the Northwestern University Robotics Club, following are eight actions required to use generative AI effectively and ethically during its early stages:

1. Keep technology honest and accurate. Many leaders have been surprised by the inaccuracy of generative AI tools as well as their “hallucinations.” The tools seemingly imagine “facts” that are not true in response to prompts and produce confident, incorrect statements. The statistical predictive nature of models mean hallucinations can occur when there is little training data relevant to a required piece of generated content. They can also occur if prompts are poorly phrased.
   
   Effective leaders coach human users to create quality prompts and verify the accuracy of content.
   
   
2. Keep technology ethical and legal. Effective leaders establish usage standards and use cases to facilitate employees’ ability to consistently respect privacy and copyrights, cite sources and only use information obtained with the creators’ consent. Because AI tools are often trained on large amounts of data, it can be difficult to source where the training data came from.
   
   Effective leaders employ techniques where content is generated from a known set of verified documents that are searched and incorporated into the context (input) of the model rather than a model containing “all” the knowledge.
   
   
3. Keep confidential information safe. Additionally, effective leaders establish usage standards that include guidelines and procedures to keep confidential company, employee and customer data from being exposed to the public. They develop and implement policies and checks to prevent proprietary information from being inadvertently released via AI platforms, including through AI learning and training. They also ensure legal and security reviews of AI services. While some services provide data security and privacy guarantees, others make it clear that users cede rights for any entered data.


4. Maintain transparency. Effective leaders inform users on how models work and educate them on their limitations. The very nature of AI generation and function means that tools can act as “black boxes,” making it difficult to accurately evaluate what the models will produce and what sources, if any, they are referencing. Effective leaders maintain transparency wherever possible for all stages of models and generated outputs. They select AI tools that list the sources (links) they have used when generating content, which helps address transparency challenges.


5. Provide context. Modern AI effectively synthesizes content on which it is trained, but it is less effective with situational awareness and analysis. While AI tools learn more each day, human users have the vital role of providing context.
   
   Effective leaders hire users into new roles such as AI prompt engineers – individuals highly skilled at leveraging generative AI tools and their output. These leaders deploy individuals in such roles to effectively use models, including information about context and desired output in a way that is both efficient for the tool and clarifying for those who use the output.
   
   
6. Provide authentic empathy, compassion and connection. Generative AI can be trained or prompted to provide output in a style that mimics empathy and compassion. Thus, AI is being used today to generate seemingly human interaction, develop “relationships” and provide emotional support.
   
   Effective leaders help employees understand that while AI models may create outputs that appear to provide emotions, they are not real. These leaders help their people interpret messages with a correct, healthy framework and set of expectations. They also provide real human connection in an increasingly digital and virtual age.
   
   
7. Address bias. Because generative AI models are trained on content that naturally includes the biases of the human users who created it, historical biases (including analytical biases such as recency bias and social biases such as discrimination) become built into models and replicated by tools as they relearn.
   
   Effective leaders operate knowing that AI is only as good as the data it is trained on. They take steps to ensure objectivity and fairness on data input and interpretation of output.
   
   
8. Complete the work. Examples of different uses of generative AI tools include writing job descriptions, creating computer code, writing sales plans, developing marketing messages, creating operations task lists, generating research, and answering routine employee and customer questions.
   
   Effective leaders understand that for some jobs, the tool may do the majority of the busywork, but people must still complete tasks by adding their insight and shaping outputs based on their skills and experience. In virtually all cases, it is up to the human user to finish the job.

Effective leaders evolve the role of human users in parallel with generative AI technology to maximize the benefits of these new and developing tools while mitigating their associated risks.

A version of this article originally appeared on Forbes on July 27, 2023.

Disclosure

One of the cited references is the son of the author.

Author

---

John M. Bremen

Managing Director and Chief Innovation & Acceleration Officer

email Email

---