[AI for Business #8] - Why most AI projects fail — and how to beat the odds
Minimising trial-and-error and accelerating AI success
Dear readers,
I’m usually quite upbeat about everything related to AI. But last week, I came across a stat that gave me pause. According to this report by RAND Corporation, over 80% of corporate AI initiatives fail.
While I’ve seen my share of AI missteps, realizing that four out of five projects don’t deliver was still hard to digest. It’s a reminder that there’s often a gap between what companies want from AI and what they’re equipped to achieve. Given the novelty of the technology for businesses, it is easy to overestimate your capabilities while underestimating the complexity of a planned AI solution. After all, we are surrounded by great marketing that makes us believe AI is easy and accessible to anyone.
To re-align, companies need to adjust their mindset and focus on building up internal AI capabilities while implementing their first projects. Binary "build-or-buy" thinking is too simplistic for AI: most teams currently don’t have the skillset to build end-to-end AI systems. On the other hand, outsourcing ("buying") everything is not a sustainable option. In my latest article, I outline an expertise-driven approach for partnering and learning, which allows organizations to upskill their internal teams while working with a trusted partner with deep AI skills.
👉 Forget build-or-buy. Instead, partner, learn, and grow.
This approach can help you:
Reduce trial-and-error in your AI journey
Align tech with actual business needs
Build internal capability while delivering results
Empower your teams for long-term autonomy
Whether you're starting out or scaling up, this article offers a practical framework to move beyond false choices and create a solid basis of AI assets and capabilities.
Read the full article: Enterprise AI: From build-or-buy to partner-and-grow
Let me know what resonates — and as always, I’d love to hear how you’re approaching these challenges in your own companies!
Warm regards,
Janna
Tales from the AI graveyard
Based on interviews with over 60 experienced AI practitioners, the report The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed outlines the patterns behind failed AI projects. Here are the most common reasons:
Failure to define the right business problem — Teams optimize for the wrong metrics or apply AI to tasks that don't need it.
Data issues — Lack of suitable, clean, or well-structured data derails many promising efforts.
Missing expertise — Organizations dive into AI without the right talent or understanding of what’s feasible.
Ok, that’s enough pessimism for today. If you’re up for something sharp and entertaining, I recommend the talk How to Fail at AI Strategy by Greg Ceccarelli and Hamel Husain. It dissects familiar missteps in corporate AI, and if you flip their advice upside down, you’ll have a collection of practices that will maximize your chances of success.
From my own experience, here are a few more general mindset shifts I believe many organizations need to make on their AI journey:
Lead with business needs, not tech trends. Start with the domain problem and bring AI in as the tool, not vice versa. IT-led projects often stay stuck in their sandbox unless grounded in real workflows and user needs, with everyone wondering why.
Respect the craft of AI. We are sold on an illusion of accessibility around AI, but in reality, it’s a complex technology. If your team doesn’t have the expertise, get it externally before you start building. Similarly, get advice if you can’t confidently estimate feasibility, effort, or ROI.
Start small and learn fast. Don’t just deliver results but build capabilities. Every AI initiative should leave your team more skilled and confident than before. Eventually, this road will lead you to real autonomy.
Enterprise AI: From build-or-buy to partner-and-grow
Tired of the build-or-buy dilemma in AI? There’s a smarter way. In my article, I introduce an expertise-driven approach to partnering in AI. Instead of asking "Should we build or buy?", start by mapping the AI and domain expertise required for each part of your AI system. This lets you identify the right collaboration model and choose partners who don’t just deliver code, but help your team learn and grow along the way.