Last year, the buzz around AI was reminiscent of the dot com boom in the late 90’s. We were standing at the edge of a new frontier, with everyone talking about how fast things were changing, the immense possibilities in front of us, and how all kinds of new commerce and businesses would be born.
That all turned out to be true…along with many things that we also didn’t see coming. These include the dot com bust that wiped out 77% of the value of the NASDAQ over a two and a half year period, the transformation of consumer preferences as Amazon grew to become the second-largest retailer in the world behind Walmart, and that people would retreat into digital spaces in such numbers that it correlates with a worldwide decline in fertility rates and intimate partnerships between men and women.
When AI hit the streets in November 2022, the power of new AI tools—ChatGPT and its siblings—led to a similar level of shock and awe. Conversations and panel discussions were filled with big, existential questions about safety, scale, and where to even begin…thanks, in part, to what we have learned since the dot com boom and bust and now a quarter century into the Internet age.
The AI landscape has changed dramatically in the relatively short period of time since we first got our hands on GPTs. At first, attention was squarely on the power of large language models (LLMs) to recall information and their ability to understand and generate text. That reminiscence seems quaint at this point.
I was recently chatting with Jeremy, a Do Good by Doing Better subscriber, about how the focus has evolved from that initial “wow” to what I'd call the industrialization of AI. I’m talking about finding and filling in all the nooks and crannies of business where AI can be put to work with the greatly expanded capabilities and functionality available in today’s AI-powered systems.
The Foundation of an AI-Enabled World
At its core, the AI stack today is made up of five core capabilities. To illustrate, I’ll employ an analogy of how these might come together to operate a pizza shop (although I’m going to stay out of the fray on whether a Hawaiian pizza is legit).
Large Language Models (LLMs). The ability of LLMs to accept natural language inputs and to respond with the same in outputs was the a major breakthrough. LLMs aren't smart in a human sense, but they've been trained on massive amounts of written material, so there’s a lot “in there.” That’s why when you ask an LLM to finish the sentence, “To make a good pizza, I need…” you might receive a list of ingredients and a recipe. This ability to take our inputs and predictively model what should come next makes them very useful for text generation, summarization, distillation, comparison, translation, and search. In our pizza analogy, LLMs provide the ability to communicate and to retrieve information necessary to craft our recipes, menus, and so on.
Toolkits. Tools provide the methods and skills to accomplish a tasks. For example, if you were building a toolkit to make pizzas, you would need methods to make the dough, roll it out, prepare ingredients, warm up the oven, check the oven temperature, and so on. In a pizza shop, you’d also need toolkits to take orders, bake pies, serve patrons, collect payment, and clean the dishes, and so on.
Agents. Agents are the orchestrators of the AI paradigm. By tasking them to follow a playbook, they use rules, data, and examples of what’s desired (and what’s not) to pursue a goal. They “understand” the task at hand by using LLMs to communicate with us and tools, and can tap into their knowledge as well. Agents can also be given access to methods and skills in each toolkit to take actions that move toward the goal it is given. In our pizza shop example, we might have an agent who is the chef, an agent to take orders, another one to seat patrons, plus a supervising agent to coordinate all the rest, with the goal being to operate a successful pizza shop.
Data: As you might have already guessed, data provides the ingredients we need to make our pizza pies, and AI applications need data that is high quality. This is where many companies will find it challenging to leverage AI at scale immediately. You can try to make something good with whatever you have sitting around, but if your ingredients (data) are not organized, clean, and fit-for-purpose, you’re likely to end up with something inedible instead of the pizza of your dreams. Kind of like the Dubious Food you get when you mix monster parts with edible foods in the Legend of Zelda video games… entirely gross and unappealing.
Surfaces: There is a race underway to create AI-infused surfaces, which I wrote about in The New Surfaces: Where AI Goes to Work. No longer is it necessary to copy-paste something into Gemini or ChatGPT only to copy-paste it out again into another application. It’s easy to upload or send model outputs to a Google Doc, use an API to access a database, and even to execute code right from within the AI interfaces themselves. While there’s an obvious productivity benefit to these integrations, the real quest is to create a new center of gravity for our attention and one that will replace the desktop metaphor of the personal computer era.
There are two reasons for this. First, the AI companies need you and your prompts, data, and attention to train their models and to improve their products. It’s why "free" AI services aren’t really free – there’s a transaction going on even if you don’t see it directly. Having all this happe in one place makes collection easier.
Second, and the long pole in the tent, is the fact that attention is the currency of the digital realm, and the AI companies are trying to peel advertising and commerce revenue away from search, social media, and commerce platforms and capture it in their own. This is the motivation driving many of the strategies we see coming into play today – to monetize your attention.
If you’d like to take a down-the-rabbit-hole journey on all the ways that technology platforms do this and the implications for free will and choice, I recommend checking out The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power by Shoshana Zuboff.
From Hype to Hard Work
You may be looking forward to the day when all the hoopla about AI will pass. Maybe you even let loose an eye roll when you read the title of this post. As a business intrapreneur who leads transformation and innovation with technology, I’ve seen it all at this point: excited wonder, blank stares, and even hostile confrontations.
But here’s the thing… You are mistaken if you think AI is a fad. What I’m seeing now is a surge of interest in building AI-enabled systems and products that solve complex, real-world challenges. The question is no longer “what can it do?” but “how can we use it to solve this specific problem?”
I firmly believe that generative AI will lead to a fundamental redesign and reordering of how work is done in the next 2-3 years. It’s a forcing function that will be used to re-engineer products, processes, and the workforce.
We’re going to have to reimagine what's required of our teams and systems, and we’ll evolve from a task-based “do this for me” mindset with AI to a paradigm where we are optimizing dynamically and in real time.
While some take an apocalyptic view of these changes, I take a practical perspective; the work to Do Good by Doing Better is about embracing a constant state of evolution and figuring out how to improve…one act at a time. After all, we constantly interact with technology and live within systems. There is an endless set of opportunities to improve customer experiences, reduce friction, consolidate data, make features like search more intuitive, and improve products. AI will transform much of this landscape, and I’d rather be a part of shaping that transformation than not.
What I hope is that we will choose to focus AI in ways that result in us having more brainpower to solve fundamental and complex problems, and to do creative work that is uniquely human. Art, inspiration, and creativity are necessary for living; human progress and innovation are not driven by algorithms—even as we use them to innovate.
Two Orders of Change
Here’s what I think we can expect to see in the next 2-3 years. The impact of AI adoption will come in two waves.
The first will hit the formulaic processes that underpin so much of our daily work. These include:
Creating boilerplate documents, first drafts of reports, and internal communications.
Automating sales lead generation, marketing campaigns, and customer support.
Simplifying data cleansing and processing.
A shift from traditional reporting and dashboards to natural-language queries for information retrieval.
If you work in one of the areas above today, I suggest you learn about AI technologies and figure out how to apply them—that way you’ll head into the transformation with new skills that are in demand in the new paradigm.
The second order of change will build off this initial wave. This is where the real disruption will come because entire categories of work will be refactored.
It will seem to start slowly but will then accelerate quickly and leave heads spinning. I've seen applications that would have taken months to build now constructed in days. While some might view this as alarming and try to stop it, many will simply dive in and use the tools available to create amazing new tools and experiences, and they will drive successes further into the marketplace. This is where the disruption will happen, especially for those who don’t engage and believe they're insulated or that their jobs are secure.
As someone who leads advanced analytics teams, up until a year or two ago, I thought there was no way that kind of work would be impacted by AI. But now I do. This shift isn't about AI being suddenly smarter than people; it's about the value proposition of some kinds of work being done by people itself being challenged.
What this means is this: If you are not adding value, you're going to be at risk, and it's going to happen very, very quickly.
An Opportunity for Creators
The good news is that I believe AI’s voracious appetite for data will create new economic opportunities for those who create novel content, bring forward new ideas, and deliver value.
It will require systemic changes to capture that value, with a first move being made by Cloudflare, which is now offering site owners the ability to block the scraping of AI content and a pay per crawl model that will let site owners charge for the use of their content for AI training.
Get Going!
The shift from being wowed by AI to putting it to work might seem daunting, but the reality is this… If you aren't actively transforming your company’s workflows and operations with AI today, you're already behind. Your competitors are likely using it and embracing AI too.
The most meaningful work we can do is to become practitioners of the new – to not wait for others to tell us what’s coming and what to do, but to lean into the discomfort of the unknown and create the future ourselves. By building our muscles of curiosity and inquiry, we can begin to see how AI can become a force for positive change in our own work and communities.
The time to get going is now, before the game changes for good.
Related Reading
Photo Bonus
Whether or not you believe that stormy skies are ahead from the effects of AI, in this case, I think we can agree that these clouds following a summertime storm were beautifully illuminated by the setting sun. I took this in one of those “glad I have my camera with me” moments that I wrote about in Being ready for the magic moment when everything clicks.