3 Reasons you're not getting the AI ROI you expected
Photo by Towfiqu barbhuiya on Unsplash
Disruptive innovation is a term that conjures images of radical change: it is new technologies sweeping aside the old, transforming industries, and reshaping the way we live and work. Yet, as history and recent experience show, the path from breakthrough to broad adoption is rarely smooth. Nowhere is this more evident than in the rise of artificial intelligence (AI). While AI is heralded as a transformative force, its real-world impact often falls short of the hype, not because the technology is lacking, but because organizations and users struggle to adapt. This post explores why disruptive innovations like AI require more than just new tools—they demand new behaviors, a strong digital foundation, and a shift in how we approach change. Drawing on lessons learned from being a change leader in technology adoption and working with organizations in their Microsoft Copilot AI Adoption, we’ll uncover what it truly takes to unlock the promise of AI and understand why you're not getting the AI return on investment (ROI) you expected.
Reason 1: You expected plug-and-play, but got behavior change
While there are many more complex definitions of disruptive innovation, I personally like the way Geoffrey A. Moore describes it in his book, Crossing the Chasm. He says discontinuous or disruptive innovations occur "any time we are introduced to products that require us to change our current mode of behavior or to modify other products and services we rely on." He goes on to say that other less-disruptive innovations are called continuous or sustaining innovations which "refers to the normal upgrading of products that does not require us to change behavior." Sustaining innovations are new models of the smart phone you are already used to using or new features in your car. A disruptive innovation would be the introduction of electric vehicles or the cellphone itself.
After two years of helping organizations adopt AI, I have come to find the root of the problem lies in expectations. AI is very clearly a disruptive innovation that requires changes in behavior and if we approached AI as such, then it would be much easier to implement and adopt. Unfortunately, due to the heavy investment from those creating AI, it has been pitched to consumers as having the ease of a sustaining innovation: they didn't want to tell people it would cost a million dollars and then another 1/2 million for training and change management, so they softened the truth. This has created a misalignment in the work needed to see a positive ROI from the tool and what organizations have the time and energy for.
Solution: Reframe AI as a behavior shift, not a feature drop.
Set the right expectations early. Position AI as a new way of working, not just a new tool. Use onboarding language that emphasizes experimentation, learning curves, and behavior change.
Design for habit formation. Integrate AI into daily workflows with nudges, prompts, and use-case-based training. Think “tiny wins” over time, not one big launch.
Celebrate behavior, not just output. Recognize and reward teams for trying AI - even if the results aren’t perfect. Normalize iteration and learning.
“If you treat AI like a microwave, you’ll be disappointed. It’s more like learning to cook.”
- Copilot
Reason 2: AI is built on top of a broken foundation
What makes the misalignment even more significant is that this disruptive innovation relies heavily on organizations having adopted the last disruptive innovation: the modern workplace. In February 2020, I was focused on helping organizations with digital transformation of their Microsoft tools. We were heavily focused on moving many users to the cloud and embracing new tools like Microsoft Teams. While some were stepping into the idea of the modern workplace and digital collaboration, many were still hesitant to move to the Cloud. One month later, when the pandemic hit, all those companies who were dragging their feet had no choice and were forced to undergo about 10 years of digital transformation in a matter of months. As a result, proper adoption of the tools never occurred. Fast-forward to five years later when we introduce AI (in my case, Microsoft's Copilot), users are not finding success because they haven’t been using the Microsoft Suite the way it was meant to: their files are not stored in SharePoint, virtual meetings may be happening in more than one platform, and some conversations may be happening outside the Cloud environment altogether. AI can be a useful tool but can only be effective when there is information to pull from.
While tools like Chat GPT (or chat-like equivalents built on the Chat GPT LLM) can be used without any work context (since you provide the context with your prompt), prompting alone has a learning curve. The early adopters and tech-savvy innovators dove right in, but the majority of users lack an understanding of effective prompting and haven't developed the habits to remember to use AI at all.
Solution: Revisit your digital hygiene before scaling AI
Audit your digital ecosystem. Are files stored where AI can access it? Do the right people have access to the right information? If not, AI won’t have the context it needs to be useful.
Close the modern workplace gap. Offer “back to basics” refreshers on cloud collaboration, file storage, and meeting practices. Make it okay to admit what never got adopted.
Transcribe meetings, even when they’re in person. Whether it’s turning on a Teams Meeting, typing out notes and storing them in the cloud, or taking a picture of handwritten notes, there are endless ways to keep your information where AI can use it.
Build bridges, not silos. Encourage cross-functional alignment on where work lives. AI thrives on shared context while fragmented systems kill its potential.
“AI isn’t magic. It’s a mirror—and it reflects the mess if we haven’t cleaned up.”
- Copilot
Reason 3: AI is tested on the wrong people
In addition to the misalignment of expectations and the failed adoption of modern work, AI has one additional fundamental issue: new technologies are usually marketed to innovators and early adopters instead of the people who will benefit the most from AI. I saw this most clearly with Microsoft's Copilot for Excel. The IT departments in charge of exploring AI targeted key groups to test out the tools and determine if they were worth the cost. In the case of Copilot for Excel, they gave it to their advanced finance teams; these were people who lived and breathed spreadsheets, which meant the only help they were interested in getting from AI was advanced data analytics, something the tool was far from capable of providing at the launch of the product. This left IT departments declaring it wasn't worth the investment. However, if they had given it to someone who didn't know how to make a pivot table and had never heard of conditional formatting, it would have changed their work life. But that didn't happen.
A similar situation occurred when the tool was tested on only the executives and senior leaders who weren't using the modern workplace the way it was meant to. When most of the pilot group have their own human assistant, they rarely see the value of a digital one.
Solution: Democratize AI by starting with the overwhelmed, not the overqualified.
Flip the pilot script. Instead of giving AI to IT and your senior leaders, start with those who have their information in the right places and have significant struggles in their workflows. They’ll see the biggest gains and give the most honest feedback.
Design for the “AI curious.” Create low-barrier entry points: “5 ways Copilot can save you 10 minutes today” or “AI for people who hate Excel.”
Measure impact by relief, not just ROI. Track time saved, confidence gained, and friction reduced, not just advanced analytics or cost savings.
“The real ROI of AI isn’t in the power users. It’s in the people who finally feel powerful.”
– Copilot
AI isn’t failing us, we’re failing to meet it where it actually lives: in behavior, context, and design. The promise of AI won’t be unlocked by more features or faster rollouts, but by rethinking how we prepare people, structure our digital environments, and choose who gets to lead the way. If we want real ROI, we have to stop treating AI like a shortcut and start treating it like a shift. The future of work isn’t about what’s possible; it’s about what we’re willing to change to make it real.
