Leveraging AI to Transform Knowledge into Action at Nurish

Introduction

In this post, I aim to discuss our experience of using AI to transform knowledge into actionable work and its implications for all of us. In today's economy, many of us are 'knowledge workers.' However, how effectively do we turn our 'knowledge' into 'work'? We partake in a variety of learning activities – reading books, attending courses, watching videos, participating in events, or exploring new ideas. While learning is enjoyable and fulfilling, I often notice a substantial disconnect between what we learn and apply, which can be frustrating.

AI's Role in Knowledge Work

I contend that AI will dramatically improve our ability to translate knowledge into practical work. AI possesses unique capabilities in processing knowledge, unlike any previous technology. Furthermore, for any given task, AI can rapidly enhance and expand our knowledge, enabling us to devise necessary actions swiftly. I am convinced that AI is well-suited to bridge our objectives with our learning and turn them into practical actions. AI goes even further by augmenting our incomplete knowledge and utilizing its training to create more practical opportunities than we could on our own. As knowledge work is the driving force behind today’s economy, the impact of this will be profound and widespread.

AI-First @ Nurish - Ideation for New Functions and Features:

I'll share my experience of using AI in a specific project at Nurish: Voice Calorie Counter . We aim to bolster Nurish with improved functions and features that persuasively encourage customers to engage more with our app. This includes trying our app, logging meals regularly, exploring different features, and making referrals. Our persuasive tactics align with our mission to assist users in making better nutrition decisions, ensuring mutual benefit.

To develop persuasive features, I turned to 'Influence: The Psychology of Persuasion' by Dr. Robert Cialdini , which details six principles of successful persuasion. My challenge was to translate these principles into practical app features. This is where AI came into play, helping to generate ideas based on Cialdini’s principles. I used Google Bard, ChatGPT, and Anthropic’s Claude, providing each with context about Nurish and our intent to add persuasive features. I then asked each AI for suggestions based on Cialdini’s principles, repeating the process with all three to maximize idea generation. I then shortlisted the ideas, refined some, and added mine. Overall, I ended up with a richer set of ideas for functions and features that I would have. Apart from the time to read the book, this process of working with the AI to generate the ideas and refine them took around 3 hrs.?

Observations from the Exercise:

• The AI's suggestions varied significantly in quality. Some ideas were excellent, others mediocre, and a few impractical.
• Interestingly, the AI proposed many practical ideas that I had yet to consider. Given my familiarity with the app, I could easily identify which of its suggestions were feasible. For instance, applying the 'Social Proof' principle, one AI suggested incorporating leaderboards to showcase desirable behaviors. While not unique, it was a practical and useful idea I had overlooked.
• The AI-generated list also sparked new and related practical ideas, including enhancements of AI suggestions.
• However, AI sometimes misinterprets key concepts. For example, regarding the principle of ‘Commitment,’ the AI overlooked the importance of creating a self-image that relates to the commitments, leading to superficial suggestions. This underscored the need for human insight to fine-tune AI-generated ideas.

I started with a particular concept and a book in mind. There are numerous approaches to using AI in knowledge work. Whether it's synthesizing research on impactful app development or analyzing top apps for user behavior insights, the paths are varied. There is no single 'right' way to utilize AI, much like conventional methods.

A View on AI for Knowledge Work

The following graph illustrates the range of ideas generated with and without AI in knowledge work. It is based on two assumptions: a) The knowledge worker focuses on a specific topic and has some foundational knowledge. b)The use of general-purpose LLMs like ChatGPT or Bard, rather than custom solutions tailored to the specific topic.

Categories of Idea Generation:

• Human-Only: Knowledgeable individuals can generate a range of high to low-quality ideas but in limited numbers.
• AI-Only: General AI tools can produce a greater number of ideas, many of which are of good quality, but they may not match human-level insight.
• AI + Human Collaboration: This combination yields the most productive results overall. The quality surpasses human-only efforts as individuals can refine AI suggestions. The quantity is less than AI-only since some AI-generated ideas are discarded.

This graph is a first step and can vary depending on the task at hand. For example, if you are looking to synthesize ideas combining macroeconomics and health outcomes, the number of human ideas can be reduced since there are not that many experts where AI might still be able to.

Implications for All of Us

The example I've shared is just a glimpse into the broader application of AI in knowledge work. Here are some thoughts based on my experience and analysis:

• AI’s application in helping book such day-to-day knowledge work is transformative, impacting all industries, roles, and levels. This is particularly true for creative work and higher-level organizational roles. The key is to start experimenting with AI now, rather than waiting for perfect tools.
• While organizations may establish policies to harness AI, that won’t be enough to help AI with different knowledge work, evolving circumstances, and innovation. individual initiative will be crucial in maximizing its potential. You cannot wait for your employer to lead you to the promised land in getting the most out of AI, they have but a role to play.?
• This shift marks a significant change in how knowledge workers will use technology. Alongside standard and specialized tools, AI will serve as a co-pilot, demanding continuous learning and adaptation, just like how software developers have to constantly upgrade their skills with new software tools.
• The AI tools available today are just the beginning. As technology and software evolve, the capabilities of AI will only improve. Those who start exploring AI now will be at the forefront of developing new work methodologies.
• Beyond improving the quality and quantity of work, AI opens new avenues for innovation, synthesizing knowledge across fields and testing multiple approaches to problems.

We are entering a new era of knowledge work. For those in established corporations, AI tools are enhancing existing roles and methods. However, AI is expanding the realm of possibilities, signaling a significant shift in the scope of work.

Conclusion

At Nurish, every experiment to apply “AI-first” brings new learning and challenges. We are sure we are not alone. We welcome your insights and experiences with AI. Feel free to share your thoughts in the comments or reach out to me directly.