Technology

AI adoption stalls despite rapid technological breakthroughs, highlighting integration challenges

Monday, 17 November 2025 5:06AM UTC

While generative AI advances exponentially, businesses grapple with practical deployment, underscoring a divide between innovation and real-world application amid organisational, infrastructural, and ethical hurdles.

Since the debut of ChatGPT in November 2022, generative AI has experienced rapid technological advancements, pushing the boundaries of large model performance and expanding multimodal capabilities. AI agents have become increasingly autonomous, capable of executing complex tasks by leveraging various tools, fueling calls about the imminent arrival of Artificial General Intelligence (AGI). However, this accelerated pace of innovation stands in stark contrast to the slower adoption and commercial implementation of AI technologies. Recent data, such as the US Ramp AI Index, reveals stagnation and even decline in the uptake of paid AI products by American companies.

A July 2025 report from the Massachusetts Institute of Technology titled "The GenAI Divide: State of AI in Business 2025" further underscores this disconnect, highlighting that 95% of generative AI projects either fail to deliver anticipated results or are abandoned midway. This finding notably unsettled US stock markets, demonstrating the substantial gap between speculative hype and commercial reality.

One of the core challenges in moving AI from intriguing technological showcases to impactful industrial applications lies in realigning business processes to effectively integrate AI capabilities. AI models currently cannot provide comprehensive end-to-end solutions; instead, businesses must strategically identify segments where AI excels, where sufficient enterprise data exists, and where value generation is highest. This approach requires segmenting workflows to allocate AI-friendly tasks to machines, while humans oversee, intervene, and manage more complex or nuanced activities, such as experience-based judgement and emotional intelligence. This process mirrors path planning in navigation, AI acts as a high-speed route for certain segments, but humans must connect and complete the overall journey. Enterprises must internalise their unique demands and constraints, map these onto the evolving capabilities of AI, and continuously revisit this alignment as technology advances.

Despite this framework, many businesses remain at a preliminary stage, deploying AI tools without dissecting workflows or systematically evaluating AI’s fit for their needs. This often results in inefficient implementations and unmet expectations.

Regarding leadership in AI deployment, two primary paths have emerged. One involves AI experts embedding themselves into industries, exemplified by the rise of Forward Deployed Engineers (FDEs) pioneered by Palantir. These engineers spend extended periods within client companies to deeply understand operational realities and identify bespoke AI applications aligned with business pain points and technical feasibility. This Silicon Valley model has gained traction and investment interest for bridging AI expertise and industrial knowhow.

The second path entails industry practitioners mastering AI tools themselves, fostering an internal culture of innovation and self-driven transformation. The "shadow AI economy," as described in the MIT report, shows that while organisational uptake of paid AI services is limited, a majority of employees use AI tools privately to enhance personal productivity. This grassroots adoption hints at latent potential for systematic integration if enterprises can coordinate these efforts and adapt AI tools to their specific contexts.

Crucially, the explosion of AI programming tools is lowering the barrier to software development, enabling domain experts without traditional coding skills to prototype and iterate AI-driven solutions rapidly using natural language interfaces. Industry leaders like Microsoft’s Satya Nadella and Google’s Sundar Pichai have noted that a significant portion of their code is now AI-generated, and experts predict that programming languages may eventually give way entirely to natural language commands. This democratization of AI development shifts the impetus for AI implementation from external technical specialists towards internal industry-driven innovation.

This trend is especially promising for small and medium-sized enterprises (SMEs), which traditionally faced hurdles in digital transformation due to fewer resources and complex legacy systems. The agility of SMEs combined with accessible AI programming tools could position them as pioneers in practical AI adoption, swiftly building tailored AI workflows without the overhead large corporations shoulder.

However, the journey to AI integration is riddled with challenges familiar across industries. Reports from Gartner and various industry analyses identify fragmented data ecosystems, unclear use cases, inadequate infrastructure, resistance to organisational change, and difficulties integrating AI with legacy systems as significant impediments. Additionally, ethical concerns and data privacy issues demand robust governance frameworks. Success in AI adoption often hinges not just on technology but on comprehensive strategies that encompass data management, technical infrastructure, organisational buy-in, and change management.

Consequently, enterprises should eschew unrealistic expectations of immediate, full-scale AI transformation. Instead, they ought to prioritise targeted initiatives where AI’s contribution is measurable and aligned with existing workflows, creating feedback loops that generate additional data for continuous AI improvement. Leveraging AI programming to rapidly test and refine these solutions can reduce costs and build internal momentum.

In the emerging AI era, professionals’ foremost assets will be vision and creativity, spotting unmet needs and deploying AI as a collaborative tool to devise superior solutions. Encouraging employees to adopt AI programming skills can catalyse workplace innovation, ultimately transforming business practices and broader industries. Though AI promises to be a potent engine of productivity and progress, it remains a partner in a co-evolutionary process, not a universal remedy.

Author Li Junjie, Managing Director of Liwa Information and Chairman of the Stanford Growth and Innovation Circle, encapsulates this nuanced reality, underscoring that AI implementation is less a dramatic upheaval and more a measured calibration between technology and industrial demand.

📌 Reference Map:

^[1] (36Kr) - Paragraphs 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11
^[2] (HP) - Paragraph 8
^[3] (Gartner) - Paragraph 1
^[4], ^[6], ^[7] (PrimeBPM, 11th.ai, ExcellentWebWorld) - Paragraph 8
^[5] (LeanIX) - Paragraph 8

Source: Noah Wire Services

More on this

https://eu.36kr.com/en/p/3556517640535175 - Please view link - unable to able to access data
https://www.hp.com/us-en/shop/tech-takes/ai-implementation-roadmap - This article discusses the complexities of AI adoption, highlighting challenges such as fragmented data ecosystems, unclear business use cases, insufficient internal expertise, and inadequate infrastructure planning. It notes that approximately 70% of AI projects fail to deliver expected business value. The piece emphasizes the need for a comprehensive AI implementation roadmap that addresses technical infrastructure requirements, data management strategies, model development approaches, and organizational change management to navigate these challenges successfully.
https://www.gartner.com/en/newsroom/press-releases/2024-09-09-gartner-predicts-40-percent-of-generative-ai-solutions-will-be-multimodal-by-2027 - Gartner predicts that by 2027, 40% of generative AI solutions will be multimodal, up from 1% in 2023. This shift from individual to multimodal models aims to enhance human-AI interaction and differentiate GenAI-enabled offerings. The article highlights the potential of multimodal GenAI to transform enterprise applications by enabling new features and functionalities, applicable across various industries and use cases.
https://www.primebpm.com/ai-for-business-process-improvement-challenges - This article outlines common challenges in implementing AI for business process improvement, including lack of clear strategy and alignment, poor data quality and fragmented systems, resistance to change, integration with legacy systems, and ethical and compliance concerns. It emphasizes the importance of addressing these obstacles to successfully integrate AI into business processes and achieve desired outcomes.
https://www.leanix.net/en/wiki/ai-governance/ai-implementation - The article discusses common challenges in AI implementation, such as data privacy and security, resistance to change, and integration with existing systems and processes. It provides actionable strategies to overcome these obstacles, including establishing robust security measures, fostering a culture of innovation and adaptability, and ensuring seamless integration with day-to-day workflows to maximize the value of AI initiatives.
https://www.eleventh.ai/blog-posts/overcoming-challenges-in-ai-implementation-2 - This piece highlights challenges in AI implementation, including integration with existing systems and processes, ethical and regulatory concerns, and the need for effective change management. It emphasizes the importance of addressing these issues to ensure successful AI adoption and integration within business operations.
https://www.excellentwebworld.com/ai-implementation-challenges/ - The article identifies top challenges in AI implementation, such as poor data quality and accessibility, resistance to change, and integration with existing infrastructure. It provides solutions to these challenges, including placing strategy at the core, ensuring cross-functional buy-in, and addressing data quality issues to enhance the success of AI projects.

Noah Fact Check Pro

The draft above was created using the information available at the time the story first emerged. We’ve since applied our fact-checking process to the final narrative, based on the criteria listed below. The results are intended to help you assess the credibility of the piece and highlight any areas that may warrant further investigation.

Freshness check

Score: 8

Notes: The narrative references a July 2025 MIT report titled 'The GenAI Divide: State of AI in Business 2025', which is corroborated by multiple reputable sources, including Forbes and Fortune. ([forbes.com](https://www.forbes.com/sites/jasonsnyder/2025/08/26/mit-finds-95-of-genai-pilots-fail-because-companies-avoid-friction/?utm_source=openai)) The article was published on November 17, 2025, indicating timely reporting.

Quotes check

Score: 7

Notes: The article includes direct quotes from the MIT report, such as '95% of generative AI application projects have poor results or are aborted midway.' Similar phrasing appears in other reports on the same study, suggesting these quotes are widely used. ([forbes.com](https://www.forbes.com/sites/jasonsnyder/2025/08/26/mit-finds-95-of-genai-pilots-fail-because-companies-avoid-friction/?utm_source=openai))

Source reliability

Score: 6

Notes: The article originates from 36Kr, a Chinese technology news outlet. While it provides detailed information, the outlet's reputation is less established compared to global media.

Plausability check

Score: 9

Notes: The claims align with findings from the MIT report, which states that 95% of generative AI projects fail to deliver anticipated results. ([forbes.com](https://www.forbes.com/sites/jasonsnyder/2025/08/26/mit-finds-95-of-genai-pilots-fail-because-companies-avoid-friction/?utm_source=openai)) The article's analysis of AI integration challenges and the need for business process reconfiguration is consistent with industry discussions.

Overall assessment

Verdict (FAIL, OPEN, PASS): PASS

Confidence (LOW, MEDIUM, HIGH): MEDIUM

Summary: The narrative presents timely and relevant information, supported by reputable sources. However, the reliance on a single, less-established outlet and the use of widely quoted phrases from the MIT report suggest a need for cautious interpretation.

AI
Generative AI
Digital Transformation