The world turned upside down in late November 2022 when OpenAI released ChatGPT, suddenly putting the power of generative artificial intelligence in everyone’s hands. It was a remarkable, breath-taking moment. Yet we could have also described the announcement another way: Uh-oh.
We were knee-deep in writing a book that emerged from interviews with more than 20 technology leaders worldwide. We intended this book to serve as a blueprint for maximizing business growth through the flexibility of application programming interfaces (APIs) – the behind-the-scenes drivers of the digital economy. As it became a labor of love, we asked ourselves: Are we missing the generative AI boat? Was our book out of date before it even reached bookshelves? Should we rethink our thesis?
After lengthy discussions, we both took deep breaths and stayed the course. That was a wise decision. “Unbundling the Enterprise: APIs, Optionality, and the Science of Happy Accidents” was published this fall, and we’re thrilled with the response. Booklife Editor named it an “Editor’s Pick,” describing the book as “an ideal balance between forward-thinking boldness and breakdowns of established best practices.”
Perhaps most gratifying is how the flood of generative AI chatbots shortly before our book’s release perfectly captured the idea of “happy accidents,” which we believe are crucial for business success in our digital world. Lo and behold, “Unbundling the Enterprise” became even more relevant because we now see how essential APIs are to the AI revolution.
Technology systems share precise slices of data through APIs. This is why they’re needed to “ground” generic intelligence models like ChatGPT with your company’s proprietary information to produce better, more trustworthy insights. APIs are also the communication method of AI agents – autonomous software that can accomplish tasks independently and are now the talk of the business world. That’s why the API economy and the AI economy are inextricably bound together.
Gartner recently named “Agentic AI” as the top strategic technology trend for 2025. But did we need the analyst firm to tell us what we already know? We can see how the generative AI train will only pick up speed in the coming months, increasing the pressure on our businesses to find measurable value from a technology that still has a wide gap between potential and reality.
With that challenge in mind, let’s unbundle how to apply three main lessons from our book as you seek ways to make AI practical within your organization.
- Optionality
- Happy Accidents
- Small Bets
Preparation beats prediction. We’re all navigating this time of incredible uncertainty and disruption. These concepts can help you think through a generative AI strategy that works best for your business.
Optionality Equals Opportunity
We opened the book with a story about two pirate captains. A grizzled old seadog named Greybeard becomes frustrated when he learns that the “X” on his map doesn’t mark the spot for buried treasure. But another modern buccaneer, Bob, appears on the scene, unloads automated digging equipment that allows him to search multiple places simultaneously, and immediately uncovers a wealth of booty. We used that metaphor to explain the importance of optionality – keeping your options open.
With technology, that means creating assets with an eye toward composability. You don’t want to make a digital process that can serve only one purpose. Instead, a modular build approach that leverages smaller, reusable components is better not only for achieving technology goals but also more innovative for financial strategies. Then, you can reassemble and configure them to create something else entirely with API connectivity.
This approach is very much like building with Legos. You’re getting more bang for your buck because you’re not constantly building things from scratch. It gives you greater flexibility. It promotes innovation because you’re always ready to move quickly when an opportunity presents itself.
Take the example of restaurants and retail outlets during the global pandemic. The ones that had invested in an API approach could quickly shift their operations to facilitate curbside delivery service. They not only survived but thrived during those unexpected headwinds. Conversely, businesses with more rigid technology backbones weren’t agile enough to adapt at a moment when innovation was everything.
In “Unbundling the Enterprise,” we also highlight other success stories – like Amazon Web Services (AWS) and Google Maps – where mega-vendors have turned core software capabilities inside organizations into APIs so they could be “called” and utilized in many unexpected ways. Transforming software into usable tools enables other smart people to envision new and profitable possibilities.
We believe that generative AI will move toward modularization, similar to the API-enabled evolution in software architecture. AI agents specializing in specific tasks will interact through APIs, requiring orchestration, interoperability, and composability. That kind of optionality will lead to unanticipated results.
Aim for Happy Accidents
These are not coincidences or luck. There’s a science to engineering them. It’s all about preparing to ensure you’re ready to seize the moment. We even created a happy accident formula that we call “OOOps.”
Happy accidents are the product of the flexibility that optionality provides in your architecture. Using value dynamics helps you quantify the best opportunities to place bets for your business. In our pirate analogy, identifying the island where to dig for buried treasure was an example of value dynamics. Business outcomes involve considering customer behavior, market trends, competitive forces, etc. Finally, there’s optimization, where you get the most from your use case experiments through feedback loops and continual refinement. Hence, OOOps. (Get it?)
For instance, we would argue that much of Facebook’s success was a happy accident because the company probably didn’t realize at the onset that collecting user data would be gold for targeted advertising.
Another example closer to home is how Boomi, which launched the cloud-based integration category, began collecting deidentified metadata from every integration in its early days. As Ed Macosky, Boomi’s chief product and technology officer, wrote: “We instinctively knew that information would be important someday, even if we didn’t know how it would be valuable.”
That fateful decision led to the Boomi Suggest wizard in the Boomi Enterprise Platform a few years later, to guide customers on how best to connect their systems. Today, the intelligence from the 300 million integrations we’ve collected fuels our Boomi AI engine to help businesses move faster.
The lesson with AI is that because it’s moving so quickly, we cannot accurately predict the future. In times of supreme disruption, it’s far better to aim for happy accidents by instead preparing for any eventuality.
Place Small Bets Widely
The rapid pace of generative AI innovation won’t slow down. The possibilities will continue bending our minds, as we don’t know where genAI is heading. But one thing we do know is that every organization is struggling to reimagine its operations and find practical value that shows ROI. What strategy will bridge the gap between this cutting-edge technology and its real-world applications?
We believe it’s better to place many smaller, inexpensive bets.
In our pirate analogy, the successful matey found gold because he was excavating in many locations. He didn’t go “all in” on one big bet. The idea of digging with 1,000 shovels, which incorporates optionality, means you stand a greater chance of hitting on something that can produce a significant payoff – a happy accident.
The added benefit is that optionality is the polar opposite of technical debt. Businesses already have to build processes and capabilities. But decomposing them into smaller chunks is a lower-cost option. That’s because loosely coupled APIs don’t require the same financial or time investment that comes with extracting capabilities from tightly contained, monolithic processes.
This approach also gives you the chance to learn more faster. The longer it takes to see the benefits of any lesson, the more expense you incur, the more time that passes, and the greater the chance you miss an opportunity. There’s less downside to losing on small bets because they don’t cost you as much, and you may still gain knowledge you can apply elsewhere.
One way the small bets concept translates to generative AI is with AI agents. You’ve probably heard about the grand vision of “agent swarms” – teams of autonomous AI agents that work together to solve complex problems. That day is coming. But today is not that day.
A happier outcome will happen by starting slowly with AI agents and tasking them to solve specific problems while keeping humans in the loop to monitor their proficiency. For instance, an agent might help you with one portion of the employee onboarding process. You learn what it can (and cannot) do and use that experience to expand toward more complicated use cases.
Combining these three ideas makes sense in the AI era:
- Orchestrate your architecture by connecting all systems to give yourself options. That will ensure access to mission-critical data while API-enabling your business capabilities.
- Spend less time trying to predict the future and more time getting ready for it. The concept of “prepare early, decide late” is valuable.
- Good things can happen when you lean into smaller, more solvable problems – even when their scope is limited.
We’re still at the art of the possible stage with genAI, but the clock is ticking to unlock value for your business. It’s time to start creating your own happy accidents.
Download “The Ultimate Guide to Practical AI,” to learn more about genAI and its everyday use cases that deliver real business value.