AI Agents: The Innovator’s Dilemma Comes for Knowledge Work
The Innovator’s Dilemma may soon apply to individual career paths
Like any good MBA graduate, I’ve come to have a few business concepts permanently etched in my brain. One of these is Clayton Christensen’s The Innovator’s Dilemma, which explores why large, established, successful businesses so often succumb to disruptive innovation in times of great technological acceleration. He argues that incumbent organizations prioritize sustainable innovation—that is, investing in making current products better for their most important customers—and in doing so, avoid truly disruptive innovations, which tend to leverage new technologies to create new markets (think streaming eating Blockbuster or the smartphone taking out Blackberry).
This is because the innovation initially targets the incumbents’ least valuable customers, who are likely being over-served, by offering an entirely new value proposition. By the time the disruptive innovation has found product market fit and successfully optimized the features that the incumbents’ high-value customers care about, the incumbent suffers from too much operational inertia and bureaucracy to challenge the disruptor. It’s not that these incumbents didn’t know these new technologies existed; but rather, they didn’t think investing in them would make sense until it was too late.
This concept is not merely business theory. Today, the acceleration of AI agents and superintelligence are bringing the Innovator’s Dilemma closer to home: our careers. The skills, mental models, and expertise that have made you successful to date might be exactly what makes you vulnerable to AI disruption.
AI Agents as Disruptors of Work
Indulge me for a moment and consider yourself the human equivalent of an incumbent organization (if it makes you feel better, I’m doing the same for myself!). Like the companies Christensen describes, you have likely been successful in your career by continuing to build upon and improve your skills within the context of the career path you’ve chosen. When faced with the option to learn a completely new professional skill or improve your ability to do what you already do, you probably often choose the latter. And this would make perfect, logical sense, until now.
The pace of industrial and technology change, though fast and always accelerating, has never been quite fast enough to disrupt knowledge work as it is today. AI agents—reasoning engines that can understand context, plan workflows, connect to external tools and data, and execute actions to achieve a defined goal—present an entirely new and different paradigm: it is us, as workers, who are being disrupted.
Agents will not just make us faster and better at executing tasks, as chatbots might, but they could also completely reimagine what it means to conduct knowledge work. Entire parts of the workforce, from entry level software developers to customer support agents to animators are watching new AI models and agents deliver their workflows almost start to finish. Such agents can complete work that is not quite as good as that of a human expert, but for orders of magnitude cheaper.
Thus, we are faced with the “good enough” innovation Christensen describes: AI agents probably won’t excel at our jobs for some time, but they could be “good enough” for the kinds of outcomes where precision is not mission-critical. The timeline to mass disruption could be shorter than we think.
Take, for example, writing software code: AI certainly isn’t as good as a tech company’s best developers, but it’s OK, and it’s faster, cheaper, and much more accessible. I have three separate friends who are in the early days of starting SaaS companies with MVP products entirely written using LLMs, and I’ve had the privilege of being along for the ride as Deloitte Engineering has continued to scale AI Assist, a tool which can at least partially automate nearly every stage of the software development lifecycle, from user story generation through to quality assurance. Of course, thankfully, humans remain critically in the loop for all of these tools, to prompt and ensure quality, ethical outcomes.
Like any disruptive innovation, once “good enough” AI agents are implemented at scale to conduct “lower-value” workflows, they could rapidly mature and demonstrate capability to deliver higher-value work in ways that are cheaper, faster, and more efficient than humans. Consider, for instance, that effective agents could be deployed 24/7, executing tasks even while we’re away from work.
The Path Forward: Taking the Leap to “Unlearn”
If I had a perfect answer for how we should reinvent ourselves in the face of lightning-fast AI innovation, I’d try my luck at the poker table. But I do propose five things to consider in the coming year:
Become comfortable with discomfort: The only thing certain about the next few years is that they are uncertain. Practice becoming comfortable with the idea that things will change fast and often. Actively seek out the ways that your work and career may change, and explore what that might mean for where you’d want to spend your time. Sometimes we will need to take a few steps back in order to leap forward, and we might as well get used to that.
Invest in adjacent skills before you need them: As you look to ways you may be able to progress in your career, consider not only the traditional path before you, but all of the horizontal and diagonal paths that may arise. If there are skills related to those you already have that are interesting to you, take the time to develop them before they become critical. The more skills you have in your arsenal, the better prepared you may be to pivot.
Get your hands dirty: Sometimes it feels like AI models and applications are moving faster than the speed of light. The only way to truly build comfort and capability with these tools is to practice. Pursue and create safe spaces to ask “stupid” questions, take risks, and make mistakes. Share your experiences with your friends and coworkers.
Strengthen your managerial skills: While the tech world likes to debate whether we will be treating AI agents as employees or tools—and it’s a fascinating debate!—there is no question that humans will play more and more of a managerial role as agents take over workflows, scoping, delegating, and reviewing work. Regardless of whether you have experience managing human teams, you can practice managing digital coworkers.
Double down on what makes you human: As I wrote about in my previous post in this blog, the one thing AI will never be quite as good at as you is being human. Your ability to be creative, empathetic, and ethical will differentiate you from your digital coworkers.
The good news is that Christensen suggests incumbents are well-positioned to disrupt themselves if they recognize the potential disruption early, take deliberate action, plan for continuous learning, and give new efforts the attention they need. By taking the initiative to think about how you can meet this moment of disruption, you can position yourself for a future you never even imagined.
-Dany Rifkin, Workforce Transformation Manager, Deloitte Consulting LLP