Lantern AI Leaders Newsletter
Two years into the generative AI surge, many enterprises have made significant investments, but results remain mixed. While the proof of concept (POC) phase may go smoothly, forward momentum often stalls before solutions are scaled or fully realized.
The issue isn’t technology. In fact, only 10% of AI roadblocks stem from AI algorithms, 20% are technology challenges. The real challenge? People, process, and organizational readiness. In this edition, we explore why so many organizations get stuck and what leaders can do to move beyond experimentation toward lasting value.
Stuck in POC Mode? Endless Experimentation Won’t Get You to Scale
A common stall point for AI initiatives is what we call POC purgatory – the cycle of building pilots that never evolve into production-scale solutions. Many organizations launch AI proofs of concept with enthusiasm, only to watch them lose momentum when it’s time to operationalize.
Why? In most cases, it’s not because the technology fails. It’s because the use case wasn’t compelling enough, the value wasn’t clearly defined, or the initiative lacked strong executive sponsorship.
As one Lantern client shared, “We built a good platform but couldn’t figure out how to get people to buy-in.”
There’s also a pattern of overly technical pilots: solutions built around what the AI can do, rather than what the business needs. These often miss the opportunity to connect to broader goals, such as efficiency gains, cost savings, or customer experience improvements. Without that business alignment, it’s difficult to generate urgency or executive buy-in.
Experimentation is essential, but without a path to value, it’s just activity, not progress.
Why AI Projects Falter Right Before the Finish Line
Even when organizations move beyond experimentation, many AI projects falter just before the finish line. These initiatives reach the pilot or deployment stage, only to be derailed by a wave of late-breaking concerns and internal hesitation.
It’s a pattern Lantern sees often. A customer service Copilot is prepped for rollout, only for stakeholders to pause and say, “But what if it’s not accurate enough?” In one case, when asked how accurate their human agents were, the client realized the AI was already outperforming them in high-volume scenarios.
This “sliding definition of done” is a common blocker. As go-live approaches, new requirements emerge: additional data sources, expanded use cases, revised accuracy thresholds. What seemed ready now feels suddenly incomplete.
Other challenges include:
- Uncertainty around cost forecasting, especially with query-based or usage-based models
- Concerns over data quality and source accuracy, particularly when AI surfaces gaps or inconsistencies in documentation that were previously overlooked
- Change management gaps, particularly around user readiness and training
These issues are about organizational trust, clarity, and confidence. And if not addressed early, they can stall even the most well-developed initiatives.
Why People Are the Real AI Roadblock
Despite rapid advancements in AI tooling and cloud infrastructure, the biggest barriers to adoption remain distinctly human. According to industry research and Lantern’s own experience, 70% of AI implementation challenges stem from people and process – not technology.
That means success depends less on whether you use GPT-4 or GPT-4 Turbo – and more on whether your organization is ready to adopt AI meaningfully.
Several people-related barriers may include:
- Lack of AI fluency among leadership: Just 8% of U.S. executives are considered AI fluent. Without a clear understanding of what AI is, what it isn’t, and how to use it strategically, leaders can’t make informed decisions or advocate effectively.
- Low trust and unclear value: If stakeholders don’t understand how AI supports their goals, they resist it or deprioritize it in favor of more familiar efforts.
- Missing enablement: Tools like Microsoft 365 Copilot are often deployed with little guidance, leading to low usage and poor ROI. Simply handing users the tech isn’t enough. They need context, training, and compelling use cases.
To overcome these challenges, organizations need more than technical deployment. They need leadership fluency, organizational alignment, and a shared language around AI value.
What You Can Do Next
If your AI efforts are stuck in experimentation or facing implementation resistance, you’re not alone. But you don’t have to stay stuck.
At Lantern, we help organizations move forward with confidence through structured leadership fluency workshops and readiness assessments designed to surface the blockers holding you back. Whether it’s aligning stakeholders, assessing maturity across your organization, or building a roadmap from pilot to scale – we can help!
Watch the full episode to dive deeper into these insights or reach out to our team to explore what AI enablement could look like in your organization.
Join us for the next AI Momentum meeting on June 24th: Register here