408 | 95% of AI Projects Fail: Here’s the Fix | Ben Tasker
Ben Tasker joins Jeff Mains to break down the real reason most AI rollouts fail: it’s almost never the technology. From his time as Dean of AI at Southern New Hampshire University to building life-saving diagnostic tools at MaineHealth, Ben has seen firsthand how change management, human-centric strategy, and data quality determine whether AI creates value or amplifies chaos. In this episode, Ben introduces the concept of the “AI Between Times” — the transition period where organizations are leaving old processes behind but haven’t yet arrived at a fully AI-integrated future. He explains why 95% of AI implementations fail, what separates the 5% that succeed, and how founders and leaders can build a culture of continuous learning that makes their organizations genuinely AI-resilient. Ben also shares a practical four-step learning cycle (Learn → Practice → Apply → Reflect) and makes the case that adaptability and flexibility — not technical expertise — are the most critical skills to develop right now. Key Takeaways [0:00] — Hook: 95% of AI implementations fail — not because of bad tech, but because of what surrounds it: poor rollouts, unengaged leaders, and broken processes that get automated at scale. [3:11] — Ben introduces the “AI Between Times” — the transition moment where organizations are leaving old processes but haven’t reached AI maturity. Change management is the most overlooked factor. [5:19] — What the 5% do differently: transparent campaigns that communicate why AI is being implemented, where, how, and what not to do — plus training pathways and upskilling resources. [7:38] — Why people resist change: 66% of individuals currently hold negative AI sentiment, driven by news around layoffs, data centers, and high-profile failures. Trust and transparency are the antidote. [7:52] — The danger of a “point solution” mindset: using AI to solve one small problem without mapping the full system. Ben’s pizza-ordering example illustrates how a narrow implementation can actually increase costs, erode trust, and kill adoption. [13:44] — AI doesn’t fix a messy organization — it amplifies it. Garbage in, garbage out. Data quality and process integrity must come before the model. [13:45] — Ben’s healthcare origin story: his first job exposed the cost of poor data infrastructure and uncommunicated change. Physicians worked around the system because no one explained the why. [15:27] — From community college to ICU: Ben’s student success algorithm took 12 months just to get faculty buy-in — a timeline the data science team never anticipated. At MaineHealth, a similar algorithm flagged ICU patients for early infection, potentially saving thousands of lives. [21:11] — Why people treat AI as just a search engine: ChatGPT reached 1 million users in 5 hours — faster than the iPhone, the computer, or the internet. Accessibility doesn’t equal strategy. Most people use 2–3 use cases and miss the full picture. [22:58] — Prompting is a skill. A prompt for a video, an image, and a report require different approaches. Without a learning plan, you’re not using AI — you’re guessing. [26:20] — AI’s primary objective is to interact with you more. The more you use it, the more personalized it becomes. But that also means being thoughtful about what data you share. [31:25] — Ben’s personal AI origin story: photographed workout equipment, had ChatGPT build CrossFit plans, then photographed his fridge for meal planning. That non-work use case built the prompting instincts he later applied professionally. [33:14] — AI augmentation is the sweet spot — not replacement. Most roles will incorporate AI; the job titles that don’t exist yet (Dean of AI, prompt engineer) are proof the economy is already shifting. [35:48] — “Prompt engineer” has at least two completely different career paths — one for marketing professionals, one for PhD-level AI researchers. Organizations need to recruit by skills, not job titles. [42:47] — What successful large-scale AI transformations have in common: leadership engagement, middle-layer exploration, and org-wide AI literacy — treated as a cultural shift, not a one-time training. [44:00] — Context is the new data. Websites, job descriptions, PowerPoint decks — these become the data sources that AI ingests. If the context isn’t clean and structured, the output won’t be either. [40:13] — The Learn → Practice → Apply → Reflect cycle: where most people break down is the reflection step. Without asking “What did I learn? Did I like it? What would I do differently?” the cycle can’t repeat effectively. [43:15] — Adaptability vs. flexibility: adaptability is accepting that change is coming; flexibility is navigating it once it arrives. These are the top two skills organizations are hiring for — and neither is a technical skill. [44:10] — Individuals who know AI earn a 52% pay premium over those who don’t. [45:07] — The one investment SaaS founders can’t skip: start using AI themselves and talk about it positively. Use your domain expertise to create context and systems that make your product AI-resilient before disruption forces the conversation. Tweetable Quotes “AI doesn’t fix a messy organization. It amplifies it.” — Ben Tasker “95% of AI implementations fail — and the 5% that succeed take care of the change management first.” — Ben Tasker “You can use it doesn’t mean you know how to use it. Just because you have access doesn’t mean you have a strategy.” — Ben Tasker “Data is the oil to the AI engine. Garbage in, garbage out — at scale.” — Ben Tasker “If leaders aren’t fully engaged with AI and using it themselves, the rollout is dead on arrival.” — Ben Tasker “AI can amplify your organization, but it can also amplify your natural abilities. The two are not the same thing.” — Ben Tasker …
408 | 95% of AI Projects Fail: Here’s the Fix | Ben Tasker Read More »
