With the rapid emergence of Generative AI models and applications, it is clear that this new technology will have an important impact on software organizations. Less clear is how organizations should think about using this technology. Use-cases, purported benefits (and risks) appear varied and increasing by the day, resulting in the temptation to wait until the dust has settled (or hype cycle smoothened) before taking action.
In these times, a helpful mental model is to think about the use of Generative AI as you would hiring an intern. There are five key dimensions that organizations should consider when using Generative AI:
- Hiring
- Instruction
- Accountability
- Oversight
- Talent Development
1. The Right Match: “Hiring” the Best Generative AI Model
In today’s landscape, there are a multitude of Large-Language Models (LLMs) and specialized applications tailored to different business needs, from code development to marketing and visual design platforms. Just as you would thoughtfully consider the scope of responsibilities for an intern, it is crucial to invest time with your team to determine the specific objectives you aim to achieve with any AI-enabled tool.
Selecting the right platform for your organization requires a comprehensive evaluation. Consider factors such as the model’s capabilities, reliability, scalability, support, and, of course, alignment with your company’s goals and values. Just like hiring an intern who matches your company’s culture and requirements, choosing the appropriate Generative AI applications is essential for seamless integration and optimal performance.
2. Enabling Success: Providing Clarity and Context to Generative AI
At Sumeru, we are fortunate to work with hugely talented interns each summer. However, these individuals do not arrive on day 1 with full contextual knowledge of our business. This requires us to provide clear instructions and iterate on tasks to fully leverage their intellectual capacity.
Similarly, when working with Generative AI, providing explicit instructions and context is crucial. Clearly defining the desired outcomes, constraints, and guidelines ensures that the model can produce outputs aligned with your organization’s objectives.
Generative AI models excel when given precise guidance, successful (and unsuccessful) examples, and active collaboration. It is important to start with clear chain of reasoning prompts (i.e., breaking down complex instructions or questions into smaller, manageable parts). By doing so, you help the model understand the logical flow of information and generate more coherent responses. For example, instead of asking a broad question like, “What are the benefits of investing in software companies?”, you can provide specific prompts like, “What are the economic advantages of software companies over traditional industries?”, “How does the scalability of software products contribute to their competitive advantage?”, etc.
Additionally, by presenting positive and negative instances, you can train the model to recognize patterns and make informed predictions. For instance, if you’re refining product marketing ideas, you should provide examples of both successful and failed content used in the past. By incorporating human expertise and feedback, you can improve the model’s performance over time. This includes prompting iterations such as “Can you explain the reasoning behind this statement?” or “What evidence supports your conclusion?” Through iterative refinement, the model should learn from its mistakes and produce more accurate and insightful responses.
3. Managerial Oversight: Ensuring Accountability in Generative AI Usage
When working with any team member, whether junior intern or AI-driven assistant, accountability ultimately rests with the Manager. Therefore, maintaining oversight and active involvement in the Generative AI process is vital. While the model can assist in generating outputs, it is essential to review and validate its suggestions, ensuring they meet your organization’s standards, requirements, and ethical considerations.
Generative AI models are tools that support decision-making but should not replace human judgment. The human operator must exercise their expertise and take responsibility for the final decisions and outcomes. By accepting accountability, you can confidently navigate the implementation of Generative AI within your software organization.
4. Safeguarding Sensitive Data: Considerations for Generative AI
Just as you wouldn’t entrust all your proprietary information to a new employee, it is critical to exercise caution when it comes to data sharing and security with any Generative AI application. Consider the sensitivity of the data involved, privacy implications, and the potential risks associated with sharing it with a third-party AI system.
To safeguard your organization’s valuable information, implement robust data protection measures such as data anonymization, encryption, and access controls. Striking the right balance between data utilization and security is crucial to maintain trust and protect your organization’s intellectual property. While many applications are being developed and significant capital deployed to provide additional security and data protection, it is important for teams to enhance their understanding of this technology in the interim.
In addition to safeguarding sensitive data, it is essential to recognize that data can serve as a long-term competitive advantage in the context of AI (where genuine proprietary data exists). The availability of proprietary data allows organizations to tailor their AI models specifically to their business needs and industry nuances. This level of customization may become essential as off-the-shelf AI solutions become more commoditized.
5. Investing in Future Success: Early Talent Development for Generative AI
While hiring and training new team members, especially those in junior roles, can be time-consuming, it is a strategic investment in developing a strong talent pipeline. Most recognize that newcomers may make mistakes and require oversight, but over time, they become productive team members, well-versed in a company’s operations and processes.
This applies to Generative AI adoption as well.
By initiating the integration of Generative AI capabilities early on, your team gains valuable experience and hones their collective AI management skills. They become more adept at understanding the strengths and limitations of Generative AI, identifying its optimal applications, and assessing when human intervention is necessary. This proactive approach ensures your company is well-positioned to leverage future developments in Generative AI and drive innovation.
Overall, adopting a mental model that compares Generative AI to working with talented junior employees can provide a valuable guide to its adoption. Just as it’s important to identify excellent talent early, we believe organization’s should bias towards thoughtful experimentation with this exciting new technology. And, of course, continue to hire the best human talent to help foster its adoption.