CIOs are increasingly abandoning in-house AI proof-of-concepts for commercial solutions due to high failure rates and low ROI on POCs. Research indicates a drop in unique AI tool development from 50% to 20%. The focus is shifting toward established vendors, aiming for smaller, achievable AI projects to incrementally build credibility and efficiency while utilizing proprietary data for potentially lucrative AI models.
In the evolving landscape of artificial intelligence, Chief Information Officers (CIOs) are increasingly abandoning in-house proof-of-concept (POC) projects in favor of commercial, off-the-shelf AI solutions. Following the disappointing outcomes of numerous AI POCs, as software vendors enhance their offerings with AI capabilities, businesses are re-evaluating their internal initiatives. By late 2023, almost half the companies surveyed by Gartner were developing unique AI tools, but this figure plummeted to around 20% by the end of 2024, as highlighted by analyst John-David Lovelock.
Despite some firms still exploring a handful of POCs, many CIOs are turning to established vendors, ranging from large language model (LLM) providers to traditional software companies integrating AI. Lovelock notes the cautious environment shaped by high POC failure rates and the heightened scrutiny over ambitious AI quests pursued since 2024. The reality check came when IDC research revealed that a staggering 88% of POCs never transitioned to broader deployment, leaving CIOs at a loss regarding the criteria for POC success.
Expert opinions reflect a shared concern over dwindling success rates for self-built AI, with Scott Wheeler from Asperitas Consulting observing that most of their clients, particularly in financial sectors, are now leaning toward off-the-shelf AI tools. He emphasizes the challenges faced by companies attempting to create their own AI models due to a lack of expertise and resources. “For most people, the juice isn’t worth the squeeze,” Wheeler states, underscoring the scarcity of skilled personnel and essential resources required to succeed.
Nevertheless, continued pressure from executives has fostered a plethora of poorly executed POCs, according to Eamonn O’Neill, CTO at Lemongrass. The initial excitement has often led to inflated expectations, as many believed AI would tackle significant challenges right away. “The attitude was, ‘Let’s get to the big hairy problems and sick AI on it,’” recalls Carmel Wynkoop from Armanino. Instead, results have primarily shown incremental improvements, requiring a shift in approach to set achievable goals.
In 2025, the market dynamics have shifted significantly, with vendors now actively promoting their AI-enhanced products to CIOs. Rather than independently sourcing or developing these tools, CIOs find themselves responding to an influx of generative AI offerings. As Lovelock observes, every software vendor is seemingly integrating AI features, sometimes without the option for companies to resist these updates.
In this new climate, companies re-evaluating their AI strategies are starting to set more realistic, smaller objectives that permit quick wins and efficiency enhancements, according to Wynkoop. This gradual approach builds momentum and credibility around AI’s value within organizations. For Wheeler, the future of AI development likely rests on tailoring AI models to leverage proprietary datasets, capable of delivering significant returns compared to the failures logged in extensive POC projects. Daniel Avancini from Indicium also views customized AI models grounded in unique data as a pathway to heightened value for companies pursuing this method.
The shift from in-house AI POCs to commercial solutions underscores a broader recognition of the pitfalls associated with self-driven AI projects. The alarming failure rates reveal the necessity for companies to adopt off-the-shelf tools, enriched by vendor expertise, rather than grapple with high-risk internal developments. Moving forward, targeting smaller, manageable AI initiatives appears to be a promising strategy for organizations eager to harness the power of AI while minimizing risk and maximizing returns.
Original Source: www.cio.com
