What’s GenAI *Really* Useful For - Update

Generative AI is still a shifting frontier, where the capabilities move fast, and only versatile technical people know precisely what the art of the technically possible is. However, it matters that more laypeople understand the domain of applicability of the new technology because the breakthroughs will come not just from the technology itself but also from its embedding into processes in organizations and the individual ways of working. People need answers to that elusively simple question.

What a difference two years make in this. As ChatGPT burst into the scene, I wrote my first impressions based on our MIT team's work on GPT-3. While not wrong, those thoughts have now given way to more sophisticated thinking.

Unfortunately, many people can not use the new tools fully because they make mistakes about what to use them for and how. There is still a lot of confusion about the use cases. This is partially because we see a long list of things, but they are not abstracted away and summarized in a few principles that people can remember.?

Many people are still confused about Generative AI's use cases. That's also because we see too many long lists of "good uses" that are hard to remember.

Based on some early work from my colleagues at MIT’s Center for Collective Intelligence, I thought of regrouping things into broad categories that apply to individuals and groups (including organizations and ecosystems). The result is helpful but still cumbersome - see below. ?

One step to simplify without losing too much signal is the following. I also like it because these processes aren't linear. They have loops and not the very list. At the very least, they have the loop represented below. You might want to start with “sense” at the top or wherever it makes sense to you.?

But I want to share a possible way to think about this that doesn’t require memorization of long lists of things - a way of generalizing into some principles.?

So - What is GenAI already good at?

Building on many others' observations, I have landed on this simple guideline:

GenAI is often useful when a horde of varied, increasingly competent, and mature but junior employees would add value to you and your teams.?

This encapsulates the current abilities and their comparison with teams of competent but constrained humans and hints at the evolving art of the possible.

Think of the following areas, with green meaning that much of this is within the art of the possible already, and yellow indicating that the jagged frontier must be treaded carefully to avoid using GenAI for things and in ways that lead to issues. ?

In short, generative AI is an obvious choice to consider for tasks where you need access to many ideas and perspectives; you are happy with some variability in the precision of the output; you want speed and scale; you can kick tires on the logic yourself; you believe the machine has enough data; you want to spend more effort than what you would be able to alone; and where ubiquitous availability across modes (text, video, sound) is an important factor. You can build a scorecard for your use cases with this.

Knowledge, ideas, and perspectives: Generative AI can gather, recombine, and present information from a wide range of sources. It synthesizes diverse viewpoints—often incorporating interdisciplinary knowledge—and can articulate them in flexible ways. This capability enables fresh insights, creative idea generation, and the ability to approach topics from multiple angles.

Precision of output: The accuracy and relevance of what Generative AI produces can vary. While it’s getting better with improved models and richer contextual prompts, it sometimes still provides responses that are off-target or factually incomplete. Regular fine-tuning and guided inputs, as well as quality data at inference time, help enhance precision.

Speed: Generative AI operates at incredible speeds, processing large amounts of information much faster than any human - or even groups of humans - could. It can scale up quickly—running multiple tasks in parallel—enabling rapid iteration and immediate feedback on creative ideas, research, or problem-solving tasks.

Transparency of logic: How and why Generative AI reaches certain conclusions or generates specific outputs remains somewhat opaque, but we are learning how to inquire into them - both with humans, and other machines.

Data available: GenAI's knowledge is largely drawn from publicly accessible datasets, ensuring it’s generally up-to-date on widely available information. However, it doesn’t have direct access to private or proprietary data unless explicitly provided, and its understanding is limited to the scope and timeliness of the data it has ingested.

Effort expendable: Humans have limited cycles to throw into anything. But AI isn't infinite, either. Generative AI can devote immense computational resources to produce more nuanced, higher-quality responses. The scale of the effort (e.g., the number of tokens processed) can, however, be dialed up as needed, allowing for deep dives, exhaustive explanations, and high-fidelity content generation when the computational budget permits.

Availability: These systems are online around the clock and can be accessed on-demand wherever there's high-speed data bandwidth and increasingly on edge devices disconnected from broadband. They don’t tire, can produce results in various formats (including text, voice, and even images), and are increasingly accessible through various interfaces and devices, making them nearly ubiquitous tools for knowledge work and creativity.

Mind the gap between theory and practice.

In closing, I want to remind all of us that this is a theoretical art of the possible. To achieve that potential, individuals must know how to manage GenAI – through personal engagement, process, and prompting (I call this AI's 3Ps).?

1. Personal, Human engagement. Ensuring that humans are still firmly in the loop and not "asleep at the wheel."
2. Process of Thought. Leading the machine through a thinking process, a sequence of requests, data inputs, etc.?
3. Prompts, one by one. Engaging optimally with the machine through better prompts (requests), one at a time.

GenAI’s most potent uses, as of today, are still those that augment us, giving us scaffoldings and exoskeletons so that we can do more, better, and faster. As we enter 2025, it is a good time to get these uses and related tools, processes, and people skills to spread their wings.?

This article is part of a series on AI-augmented Collective Intelligence and the organizational, process, and skill infrastructure design that delivers the best performance for today's organizations. More here and in the white paper here.

Get in touch if you want these capabilities to augment your organization. Build learning, problem-solving, innovative, intelligent organizations. Build Superminds.