The AI Prioritization Reality Check

88% of organizations now use AI. Only 39% see bottom-line impact. The difference isn't technology it's knowing what not to build. This week, we're examining the six red flags that predict AI failure before you invest, and why the most effective leaders don't build more initiatives—they filter better.

You've probably seen this pattern: Companies rush to adopt AI because "everyone else is," commit millions in budget, launch pilots with fanfare... and 18 months later, they're explaining to the board why none of it scaled.

42% of companies abandoned AI initiatives in 2025, up from 17% in 2024. The gap between adoption and actual impact has never been wider.

Following this week’s deep dive on The AI Prioritization Framework: How to Know Where to Invest First, this edition focuses on the other side of strategic clarity:

👉 Knowing when not to invest in AI—yet.

Because the most effective AI leaders don’t build more.

They filter better.

Knowing what not to build is as critical as knowing what to prioritize. With plethora of stalled or abandoned AI programs, a clear pattern emerges:

You read that a competitor launched an AI feature. You see analyst reports saying your category needs AI. A board member asks, "What's our AI strategy?"

So you commit resources to an AI project... without validating that customers want it or that it solves a real problem you have.

The test: If you can't articulate the specific business problem this solves without using the words "AI," "ML," or "innovation," you don't have a real initiative. You have a press release disguised as a product.

The deeper problem: Shiny object syndrome. If it doesn't connect to a core business metric revenue, retention, cost reduction, customer satisfaction pause it. AI for AI's sake becomes a demo reel, not a business driver.

Better approach: Start with the pain point. Then evaluate if AI is actually the right solution or if you just want it to be.

Your team proposes an exciting AI initiative. In month three of planning, someone asks: "Do we have the training data?"

The answer is some version of: "Not yet, but we can collect it" or "It's in our legacy systems, we just need to extract it" or "We'll need to label it, but that shouldn't take long."

What this actually means: Your AI project has a 6-12 month data infrastructure project hidden inside it. 43% of organizations cite data quality and readiness as their top AI obstacle.³

The fatal assumption: "The data will improve later." It won't. Weak data poisons trust early. If your first AI system delivers unreliable results because of poor data, you've just created organizational skepticism that will haunt every future AI initiative.

The test: Can you produce a representative sample dataset in 48 hours? If not, you don't have an AI project…you have a data project with an AI project hopefully at the end.

Better approach: Either start with the data infrastructure project (and be honest about the timeline), or pivot to an AI initiative where the data already exists and is accessible.

The initiative looks great on paper. The business case is solid. The technology is feasible. But when you ask "Who's accountable for making sure this gets adopted and delivers value?" you get vague answers.

Marketing thinks Product owns it. Product thinks Engineering owns it. Engineering built it and considers their job done.

The reality: AI without accountability becomes a demo, not a system. Someone needs to own not just the build, but the adoption, the iteration based on feedback, the measurement of business impact, and the organizational change required to make it work.

The test: Can you name the single person who will lose sleep if this AI initiative fails to deliver business value? If there's no clear decision owner, you're building a science project.

Better approach: Assign an executive sponsor before you write a line of code. Their job isn't just approval it's removing blockers, driving adoption, and ensuring the initiative connects to business outcomes.

The initiative requires:

...all at once.

This is why over 80% of AI projects fail.⁴ Not because any individual element is impossible, but because you're attempting to learn everything simultaneously while delivering business value.

The capability mismatch problem: If your team can't operate or explain it, adoption will stall. An AI system that requires a PhD to interpret or specialized skills to maintain becomes shelf-ware the moment your contractors leave or your data scientist quits.

The test: Can you identify at least 50% of the required capabilities that your team has already demonstrated? If not, this isn't a strategic first move, it's a gamble.

Better approach: Break the moonshot into phases. Tackle the capability-building elements in lower-stakes projects first, then execute the complex initiative when you have proven competence in the foundational pieces.

Your AI initiative requires sales to change their workflow, give up their spreadsheets, and trust an algorithm's recommendations. Or it needs customer service to rely on AI triage instead of their intuition. Or it demands that executives make decisions based on ML predictions rather than gut feel.

And your plan for adoption is... training videos and a launch email.

McKinsey research consistently shows that culture and workflow redesign, not technology, determine AI success.⁵ The companies that achieve significant value are three times more likely to have senior leaders who actively champion AI adoption and role model its use.

The test: Have you identified executive sponsors who will use the system themselves? Do you have champions within the affected teams who want this tool? Is there a clear answer to "What's in it for me?" for end users?

If your change management plan is "they'll see the value once they try it," you're headed for an expensive pilot that gets great internal demos and zero actual adoption.

Better approach: Start with initiatives where users are already asking for AI assistance, or where the status quo is so painful that people are actively seeking solutions. Build momentum with willing adopters before attempting to change entrenched workflows.

You've defined success as "deploy an AI system" or "improve efficiency" or "enhance customer experience." These aren't success metrics, they're activity metrics masquerading as outcomes.

Only about 39% of McKinsey survey respondents report enterprise-level EBIT impact from AI, despite 88% of organizations using AI in at least one function.² The gap? Lack of clear value metrics tied to business performance.

The test: Can you quantify, in specific numbers, what success looks like? "Reduce customer service handle time by 15% for Tier 1 inquiries" is measurable. "Make customer service better with AI" is not.

Better approach: Define success metrics before building anything. If you can't measure impact, you can't prove value and your AI program will die in the next budget cycle when someone asks "What did we actually get from this?"

Notice how these red flags connect to our four dimensions mentioned in the blog this week The AI Prioritization Framework: How to Know Where to Invest First

Red Flags #1 & #6 → Weak Business Impact (building without clear value)

Red Flags #2 & #4 → High Implementation Complexity (underestimating technical requirements)

Red Flags #3 & #5 → Poor Organizational Readiness (no ownership, no adoption plan)

Red Flag #4 → Negative Capability Building (trying to learn everything at once)

"More Human: How the Power of AI Can Transform the Way You Lead"

by Rasmus Hougaard & Jacqueline Carter (Harvard Business Review Press, March 2025)

Why it matters: While most books focus on AI's technical capabilities, Hougaard and Carter tackle the question every leader is wrestling with: How do I lead effectively when AI handles more of the work?

The authors, founders of Potential Project with two decades of research on leadership mindfulness, present a compelling counter-argument to the "AI will replace humans" narrative. Instead, they argue that AI's rise demands more human leadership, not less. The leaders who thrive will be those who use AI to enhance their awareness, wisdom, and compassion rather than abdicate these qualities to automation.

Key insight: "AI has the potential to transform leadership and business or to lead us toward an automated and uninspiring work experience. Which will it be?" The book provides practical frameworks for maintaining human-centered leadership while scaling AI capabilities.

My take: This is essential reading if you're implementing any of the AI initiatives we've discussed. The technical execution is only half the challenge. The real work is preserving the human elements that drive trust, creativity, and sustained organizational change. Hougaard and Carter give you the roadmap for doing both simultaneously.

→ Available on Amazon | Audiobook: 9 hourshours

"The AI in Business Podcast" hosted by Daniel Faggella (Emerj AI Research)

The gist: This isn't another AI hype podcast. Faggella interviews senior executives from Fortune 2000 companies and unicorn startups who are actually deploying AI at scale—discussing what worked, what failed, and the real ROI they're seeing.

Why I'm sharing it: If you're a non-technical business leader trying to translate AI capabilities into business strategy, this podcast bridges that gap brilliantly. Recent episodes cover topics like identifying AI opportunities that align with business strategy, managing AI pilot programs, and measuring actual returns on AI investments.

The conversations are refreshingly honest. Guests discuss failures alongside successes, providing the kind of candid insights you won't find in vendor presentations or conference keynotes. Each episode delivers tactical frameworks you can apply immediately, perfect for the 30-minute commute.

Best recent episode: "Why 95% of AI Pilots Fail" featuring enterprise leaders discussing what separates the 5% that scale from the majority that stall.

→ Listen on Spotify | Apple Podcasts | Episodes: 30-45 min

Question for you: Which of these six red flags have you encountered in your organization?

Have you seen AI initiatives stall because of poor data readiness? Lack of executive ownership? Or maybe you've successfully navigated past these obstacles, what made the difference?

Hit reply and share your experience. I'm gathering insights for a follow-up piece on the patterns that distinguish companies that successfully scale AI from those stuck in pilot purgatory.

1. "The State of AI in 2025: Agents, Innovation, and Transformation" – McKinsey

McKinsey's latest survey of 1,993 participants across 105 countries reveals the stark gap between AI adoption (88%) and AI value capture (39%). Essential reading for understanding why adoption alone doesn't equal success.

→ Read the full report

2. "Data Quality is Not Being Prioritized on AI Projects" – Qlik

Eye-opening survey showing that 96% of data professionals expect poor AI data quality to cause major crises, yet leadership remains focused on AI investment over data infrastructure. The disconnect that's killing AI ROI.

→ Read the analysis

3. "Why Most Enterprise AI Projects Fail—And the Patterns That Actually Work" – WorkOS

Detailed breakdown of the four patterns separating winners from failures: pilot paralysis, model fetishism, disconnected tribes, and data debt. Includes real case studies from Lumen Technologies ($50M annual savings) and Air India (4M queries automated).

→ Read the case studies

🫡Until next time, stay courageous, stay visionary, and keep building the future you believe in.

LinkedIn | Medium | The Leap Weekly