88% of AI projects never leave the lab

In an exclusive interview with CIO&Leader, Lenovo’s Asia Pacific Lead Amith Parameshwara tackles the reality behind AI adoption challenges.

In the rush to embrace artificial intelligence, organizations are discovering a sobering reality: implementation is far more challenging than the hype suggests. A striking statistic reveals that only 12% of AI proof of concepts ever makes it to production deployment. In this candid conversation with CIO&Leader, Amith Parameshwara, Lenovo’s Asia Pacific Lead, cuts through the AI buzzwords to address this concerning trend.

Drawing from Lenovo’s comprehensive 2025 AI survey, Parameshwara reveals how enterprise AI investments are set to triple in the next year, with 42% flowing specifically to generative AI applications. But as budgets shift dramatically, so do expectations—CIOs now face mounting pressure to demonstrate concrete business outcomes and dollar-value returns rather than just technological achievements.

Parameshwara unpacks the five critical bottlenecks preventing successful AI implementation—from persistent data quality issues to the complex challenges of scaling beyond controlled experiments. He also introduces Lenovo’s strategic Pocket2Cloud portfolio and AI Center of Excellence, designed to guide organizations through the treacherous journey from AI concept to business transformation.

CIO&Leader: How do you see CIOs’ pain points regarding AI adoption, and what solutions does Lenovo offer?

Amith Parameshwara: There’s considerable hype around AI, but organizations struggle with realizing its value. Our recent 2025 AI survey revealed three key findings:

1. Investment in AI will triple as a percentage of total technology investment in the next 12 months, which represents a significant shift in budget allocation.
2. Expectations are shifting from demonstrating technology outcomes to business outcomes. CIOs are now under pressure to show business value, ROI, and actual dollar value from their AI investments.
3. Nearly 42% of AI investment is expected to go to generative AI, despite this technology only being mainstream for the past two to three years.

Our survey identified five major bottlenecks executives face in AI implementation:

1. Data quality issues – Despite years of discussion, this remains a significant challenge. Organizations still struggle with data accessibility, consistency, and governance.
2. Infrastructure and technology costs – Optimizing costs while ensuring a proper, scalable foundation is a delicate balancing act. AI workloads require specialized hardware that can be expensive to acquire and maintain.
3. Integration challenges – Incorporating AI into existing systems and processes is crucial for value realization and adoption. Without seamless integration, AI remains siloed and ineffective.
4. Scaling difficulties – Moving from proof of concept to organization-wide implementation requires different approaches to both technology and change management.
5. Endpoint deployment concerns – Ensuring AI is delivered to the point of need across various users, whether that’s on mobile devices or manufacturing floors.

Our Pocket2Cloud portfolio addresses these challenges with:

• AI-optimized infrastructure across servers, storage, workstations, and personal computers designed specifically for AI workloads
• Platforms and solutions for managing data and developing/deploying AI, including specialized tools for enterprise-grade generative AI
• Comprehensive services spanning the entire engagement lifecycle from advisory to development, deployment, and ongoing management

Our offerings help customers balance quick wins with sustained transformation, innovation with appropriate risk management, and testing with operational scaling. We’re also integrating AI capabilities into our existing solutions to enhance their functionality and value.

CIO&Leader: How has Lenovo’s AI Center of Excellence been launched, and what role will it play in CIOs’ transformation journeys?

Amith Parameshwara: Our research shows technology executives want to focus on business outcomes and drive business transformations. While Lenovo has all the elements needed for this transformation, they existed in multiple divisions. We wanted to bring everything together with a One Lenovo approach, focusing on solving critical business problems for our customers.

The AI Center of Excellence is housed within our Solutions and Services Group with a core mandate to drive business outcomes. We’ve assembled technology experts, AI generalists, industry specialists, and professionals with skills in data modernization and change management—all essential components for successful AI transformation.

We partner with customers throughout their journey:

• Identifying critical business outcomes through collaborative sessions
• Conducting readiness workshops to establish clear objectives
• Assessing their preparedness using our comprehensive framework that evaluates security, people, technology, and processes
• Identifying gaps and creating AI adoption roadmaps tailored to each organization’s unique situation
• Co-creating AI solutions tailored to their needs rather than imposing predetermined solutions
• Supporting them from development through deployment and management with hands-on guidance

This comprehensive approach ensures that our customers don’t simply implement technology, but achieve meaningful business transformation with measurable results.

CIO&Leader: How can Lenovo help overcome skills challenges as organizations move from pilot to production?

Amith Parameshwara: Only 12% of AI proof of concepts make it to production, which is a concerning statistic. While innovation naturally has a high failure rate, this percentage is particularly low. There are several key reasons for this low success rate:

• Choosing the wrong problems to solve – Many organizations select problems that aren’t critical or aligned with their strategic vision. They pursue AI applications that are trendy rather than transformative. Lenovo helps through workshops that identify priorities and areas where AI can add genuine value.
• Lack of executive sponsorship – Without a strong sponsor at the executive level, scaling AI across the organization becomes nearly impossible. Middle management initiatives often fail without top-level support because they lack the organizational influence to overcome resistance.
• Inadequate technology foundations – Creating a small proof of concept in a lab is vastly different from scaling across an organization, which requires robust technical infrastructure including computing resources, networking, and storage optimized for AI workloads.
• Data availability issues – POCs need only small data subsets, but organization-wide deployment requires diverse data that’s often scattered across legacy systems or even in Excel spreadsheets. Data integration becomes a major hurdle that wasn’t apparent during the POC phase.
• Insufficient change management – Value realization occurs only when people actually use the solution. Integrating AI into existing processes and motivating users is critical but often overlooked. The human element of AI adoption is frequently the most challenging.

We help organizations establish the right foundations from the beginning, ensuring executive buy-in, building scalable infrastructure, addressing data integration issues, and implementing effective change management strategies.

CIO&Leader: Do business leaders demand ROI metrics at the beginning, or should they first establish business objectives?

Amith Parameshwara: The funding landscape has shifted significantly. While data science teams, CIOs, and CDOs traditionally sponsored AI initiatives, much of the budget now comes from functional business units. For example, our AI-based marketing content creation solution is sponsored by CMOs who have the budget and will benefit directly from the technology.

Regarding ROI, there are two dimensions: how to measure it and how to ensure it’s positive. Creating the right business case from the beginning is crucial—organizations shouldn’t initiate AI work without this foundation. Later regrets are common when this step is skipped.

Value realization happens only when:

1. AI is scaled across the organization
2. End users actually adopt the solution
3. The system is integrated into existing processes
4. The AI outputs lead to actionable insights

For example, we implemented a computer vision-based defect detection system for Lotus Automobiles, a high-end sports car manufacturer. We established metrics before starting the project, resulting in measurable outcomes: 80% reduction in installation errors and 99% better defect detection rates. This exemplifies the importance of defining success metrics from the outset.

CIO&Leader: How do you collaborate with the ecosystem, and what advice do you have for CIOs on AI adoption?

Amith Parameshwara: AI is a complex combination of multiple technologies requiring collaboration among various players. A strong ecosystem is the cornerstone of our AI strategy. We partner with OEMs, solution providers, technology providers, consulting companies, industry bodies, and policy consortiums to create holistic solutions.

We also invest in AI startups through our Lenovo Capital Incubator Group to incorporate cutting-edge innovations into our portfolio. This allows us to bring emerging technologies to our enterprise customers in a validated, enterprise-ready format.

Our ISV ecosystem includes over 50 solution providers who have created more than 165 AI solutions optimized for Lenovo infrastructure across multiple industries, covering both horizontal and industry-specific applications.

NVIDIA exemplifies our partnership approach. We collaborate on joint offerings like GPU-as-a-Service, providing flexibility for our customers to use GPU infrastructure as needed, and Fast Start for creating quick proof of concepts. This partnership spans joint offering creation, go-to-market strategies, and service delivery.

Looking at the next 3-5 years, technology adoption will evolve along two dimensions:

1. IT for AI: Building and managing infrastructure to support AI growth. Our research found 63% of organizations are looking at hybrid AI infrastructure across cloud, on-premises, and private cloud. AI PCs will be crucial, with 90% of organizations planning to adopt them. Our AI Now personal assistant and xCloud offering help manage growing AI workloads.
2. AI for IT: Leveraging AI to improve IT operations through AI-based sustainability solutions, cyber-resilience tools, and generative AI for knowledge management and customer support. As AI workloads grow, sustainability becomes increasingly important, and our solutions help manage data centers in an environmentally responsible way.