Organizing for innovation

Making Decisions in High Uncertainty

When working on a Generative AI or other innovation projects, the sheer volume of decisions can be overwhelming, especially when uncertainty is high. It's impossible to do everything, so how do you prioritize? How do you determine what to focus on and what to leave out?

Challenges in Generative AI Projects

In projects involving Generative AI, decision-making becomes even more daunting. Teams often feel confused and hesitant due to the multitude of unknowns and constant changes. For instance, consider MS Copilot: in January, it required a $100k investment, but now small pilots are feasible. Should you decide now, or wait for potentially cheaper options and improved technology in the near future? How do you make decisions in high uncertainty?

Choosing Between Solutions

The first dilemma arises when choosing between a generic GenAI solution like MS Copilot or investing in domain-specific language models. What would be best for your organization?

A Different Approach to Decision-Making

In navigating these highly uncertain circumstances, it's crucial to adopt an alternative approach to decision-making. While there's no one-size-fits-all solution, there are certainly better and worse ways to make decisions regarding Generative AI projects.

Start with Problem Identification

Begin by identifying the problem you aim to solve and its urgency. For example, if your firm is currently facing $50k per month in productivity losses, immediate action is warranted. Even if a small pilot could have been cheaper in the future, paying $100k for one in December was justifiable.

Focusing on the problem to solve, will probably also help you to know whether a generic or more specific Generative AI solution will be better.

Starting Small for Better Progress

In practice, starting small has proven successful for teams facing high uncertainty. This approach facilitates significant progress by allowing teams to learn what works and what doesn't. Moreover, starting small enables adaptability to changes in the environment, facilitating quicker movement through the cycles of OODA loops (Observe, Orient, Decide, Act), recommended for rapid decision-making in dynamic situations. The faster you can learn, the sooner you will know if you took the right decision(s).

Avoid Traditional Risk Management Strategies

Traditional risk management strategies may not be effective in uncertain environments lacking reliable data. Plans such as scenario planning, risk assessment matrices, or decision trees offer no additional certainty when data is scarce.

Consider Stakeholder Impact Wisely

While considering the impact of decisions on stakeholders is crucial, in the context of Generative AI, their close involvement may exacerbate confusion. It's essential to keep stakeholders informed while being prudent in sharing information to avoid unnecessary confusion. Involving stakeholders in a rollercoaster ride of high uncertainty is not advisable, as such experiences are seldom appreciated in professional settings. So choose wisely what information to share, less may be better than more.

Break Circular Dependencies

Avoid the trap of circular dependencies where decisions are interdependent, hindering progress. To break such loops, make an assumption that enables action, then test it quickly to determine whether to continue or pivot.

Define Waiting Parameters

Waiting to make a decision can be beneficial if waiting parameters are clearly defined. Simply hoping for circumstances to improve without a clear plan may lead to falling behind competitors who take decisive action.

Tips to Aid Decision-Making

Realize that while best practices aren't always applicable, they're often reliable guides. Additionally, few decisions are irreversible, so opting for reversibility whenever possible is prudent. When decisions are irreversible, follow the affordable loss principle to mitigate risks.

Embrace Data Collection and Experimentation

Finally, collect data whenever possible, and in the absence of data, run experiments to generate the necessary information for significant decisions.

Conclusion

These suggestions aim to assist in navigating decision-making in high uncertainty, offering a framework for making informed choices despite challenging circumstances.

Success with making all the decisions to bring your Generative AI project(s) to a successful end!

Subscribe to our Newsletter

Did you like what you read? Sign up to our monthly newsletter and receive our blogs and other news updates in your inbox!