5 Strategies to Move AI from Pilot to Production, According to Dell
Meet the Authors
Key Takeaways
95% of AI pilots fail to produce measurable business impact, highlighting a need for strategic shifts in design and deployment to avoid costly experiments.
Key reasons for high failure rates include lack of executive sponsorship, unclear workflow integration, and ineffective change management; best practices involve aligning pilots with core business priorities and conducting thorough workflow mapping.
Data fragmentation and reliance on synthetic data pose significant risks; organizations should utilize real data and consider a data lakehouse approach to integrate diverse data sources effectively.
As enterprise investment in artificial intelligence reaches unprecedented levels, a sobering statistic has emerged to challenge IT leaders: 95% of AI pilots fail to produce measurable business impact. This figure, highlighted in a recent MIT report cited by Dell Technologies, points to a crisis in execution rather than innovation. For SAP professionals and IT executives navigating complex ERP landscapes, the message is clear: without a strategic shift in how pilots are designed and deployed, most AI initiatives will remain expensive experiments.
According to Ricardo Calderon, Vice President of Marketing for Services & Payment Solutions at Dell Technologies, the primary reasons for this high failure rate include a lack of executive sponsorship, unclear workflow integration, and challenging change management. To help organizations navigate these hurdles, Dell has outlined five best practices designed to validate business feasibility and unlock ROI.
- Focus Pilots on Core Business Priorities
The first step to avoiding failure is alignment. IT cannot select AI applications in a vacuum; they must engage business stakeholders directly. Calderon advises pilots must align with core business functions. For SAP customers, this means targeting high-value processes within Finance, Supply Chain, or HR, rather than focusing on peripheral systems.
Explore related questions
- Map Workflows Before Applying Technology
Doug Schmitt, President of Dell Technologies Services, warns, “applying AI to bad processes will greatly hinder ROI.” This is critical for SAP organizations migrating to SAP S/4HANA, where clean core initiatives are paramount. Before deploying generative AI, Dell urges “value stream mapping” to define workflows and identify specific pain points. Schmitt notes Dell’s services organization digitized over 4,000 business processes to define the “happy path” of customer satisfaction before applying machine learning tools for automation.
- Break Down Data Silos with a Data Lakehouse
Data fragmentation between the ERP core and external cloud applications remains a major hurdle for SAP environments. While developers often default to the public cloud, integrating enterprise data there is time-consuming and risks data boundary concerns. A more effective approach is using a data lakehouse to quickly integrate cloud and enterprise data, breaking silos and simplifying preparation.
- Use Real Data, Not Synthetic
Relying on synthetic data during the pilot phase risks delays later when full enterprise data integration becomes necessary. Dell Services advocates for a secure approach that integrates clean, real customer data immediately, providing a realistic view of system performance from the start.
- Prioritize Change Management
The MIT report cited by Dell names “challenging change management” as a top cause of AI failure. Since AI projects dramatically shift operations, Calderon argues change management is crucial. He suggests utilizing tools for workforce sentiment analysis to help IT organizations succeed in new technology rollouts.
The Case for On-Premises Efficiency
The financial argument for rethinking the cloud-first mentality associated with modernizing SAP environments is compelling. A recent study, commissioned by Dell, found that inferencing and fine-tuning with the Dell AI Factory on-premises is up to 63% more cost-effective than public cloud solutions over four years.
As organizations move from strategy to scale, reliance on external expertise grows. Internal research conducted by Dell shows a significant majority of IT decision-makers prioritize service providers with end-to-end AI capabilities—spanning strategy, data prep, and deployment—to drive consistent ROI.
What This Means for SAPinsiders
AI governance now requires a data supply chain focus. Leaders must shift from managing uptime to managing data supply chains for AI models. Evaluate vendors on their end-to-end data services that prevent fragmentation, not just software features. Prioritize partners who align business goals with IT infrastructure for maximum agility.
Your operations will increasingly focus on data hygiene. Since 91% of organizations struggle with preparing data for AI, solve this bottleneck first. Investigate on-premise data lakehouses to reduce latency and avoid cloud data boundary risks. Secure, on-premise sandboxes for testing real data will become a standard requirement for development teams.
Successful ERP integration requires a massive cultural shift. Actively engage business stakeholders to map “happy path” workflows before automating them with generative models. Ignoring workforce sentiment may result in your project joining the 95% of pilots that fail to launch. Leverage external expertise for change management to ensure teams utilize deployed tools.
