Deeper Dive into the Value of AI (Part 2)

If you have been following along, I’ve been writing about the ways that AI can deliver value for an organization. My last post (https://www.dhirubhai.net/pulse/deeper-dive-value-ai-part-1-adam-krob-ecpyc/) focused on two sources of value – easing simple tasks and simplifying complex ones. These sources of AI value broadly resonate with the way we have thought of technology providing value for decades. The two that I am diving into today – acting as a thought partner and becoming the interface – are very different, both in the capabilities that AI provides us and the potential value that it can deliver. If you haven’t read my first article that reviews all four, check out – https://www.dhirubhai.net/pulse/youre-starting-use-ai-your-organization-how-you-know-its-adam-krob-jngbc/.

Acting as a thought partner

The first area of value for AI that I want to discuss today is AI as a thought partner. AI as a thought partner has emerged as large language models (LLMs) have gotten more reliable and powerful. One of the ways you can start a great conversation with an LLM is to give it a role to play. The LLM will answer questions (and pose questions to you) playing a specific role – a colleague, a superior, or an industry peer. Geoff Woods, the author of AI Leadership, discusses this source of value by using a very powerful analogy. As the owner of his own company, Woods requires the advice of board members daily for communications, planning, or policy. Most board members don’t sign up for this level of commitment. Woods has set up a custom chat with several pre-defined personas to give him advice as a virtual board. If he has an important decision, he runs it by his virtual board first.

Why is AI capability as a thought partner so important? There are two important reasons. The first is that we are a product of our experiences and our choice (ok, and our genes). This set of experiences is a key source of our value but is also a limitation. Marshall Goldsmith’s What Got You Here Won’t Get You There is a great source for explaining how our greatest strengths often hold us back. I remember several people who were promoted in IT organizations that had tremendous technical skills but had neither the desire nor the ability to be a manager. The second is that, as people move higher in an organization, they have fewer peers. This problem becomes acute at senior leadership levels where many executives report a sense of loneliness. That certainly explains the comforting feeling you get at a conference when you find out that everyone else has the same problems you do!

LLMs step in because of how they are trained. Base models are trained on the stories that everyone tells. They have read a thousand articles about failed ERP implementations and several thousand articles on reducing insurance costs for public works projects. Those stories provide a widely varied set of experiences and choices to give feedback on our projects, but also to challenge our own assumptions (based on who we are).

I would suggest that this is harder value to quantify, but if a thought partner can help us derisk projects, find out where our blind spots are, and give us more consistent feedback from expert perspectives.

Becoming the interface

The last source of value is the one that excites me the most. Since technology became a core component of how we work, our activities have become increasingly dependent on the way technology requires us to work. My favorite analogy is Henry Ford expecting humans to behave like a cog in the assembly line machine. Technology requires that we interact with it in the way it wants to be interacted with. How many ERP systems are driven by a strictly defined interface where human actions are constrained by the create, update, and delete functions of the back-end database.

Interface design has come a long way – think of the difference between text-based games like Oregon Trail and the latest Star Wars game on the PS5 – but it still is limited (and limiting). We are being forced to act like a computer, not like a human. Humans are story-telling creatures. We experience the world as an unfolding set of choices (a choose-your-own-adventure book) and relate our history that way.

LLMs, on the other hand, offer an infinitely flexible interface between humans and their technology. Let’s talk first about the reason why. LLMs are trained on stories. It’s how they began and the mode that they are most comfortable with. LLMs can then summarize the data from the story and provide it to the technological back end. One of the greatest advantages of this approach to an interface is that order doesn’t matter. Humans often eventually “get to the point” and the LLM will patiently converse with them until they provide what the back-end technology needs (or ask follow-up questions to make sure that the database has everything it needs).

Additionally, LLMs can help us interrogate information. We don’t have to be an expert in Excel or R or SQL to make sense of a data set. We can ask the AI to find outliers and highlight them or explain model fit without using the term R-squared at all

LLMs can democratize the analysis of information, putting tools into the hands of subject matter experts with no training in data analytics. It allows us to marry the data to the people best able to interpret it, without the need for training or the limitations of what an interface designer imagined they might want to do.

A final thought on AI value

We are just beginning to uncover the value that AI will bring to our organizations. I’m sure that, five years from now, the value that AI is delivering is different than what I anticipated. We are at the beginning of a transformation in how we interact with technology. If we embrace it, we will derive huge value for ourselves and our companies. If we don’t, we will watch others pull further ahead.