Many AI initiatives fail due to a lack of data readiness, with data trapped in silos or riddled with errors. Ryan McElroy, Vice President of Technology at Hylaine, a values-first technology consulting firm, shares insights on overcoming these challenges in a recent Q&A. McElroy identifies five common roadblocks: data access, siloed systems, data quality, governance, and the human factor. He notes that data access issues often stem from legal or security blocks or incompatible formats, while siloed data persists as enterprises operate across multiple cloud platforms. Quality problems like inaccuracies and redundancies undermine model accuracy, and governance adds complexity in ensuring compliance with privacy regulations, particularly in sensitive industries like insurance and healthcare.
To ensure AI success, McElroy advises tech leaders to build a mature, AI-ready data infrastructure. This includes investing in data engineering tools and modernizing data architectures to handle the scale and velocity AI requires. He highlights the importance of data reliability engineering (DRE) as a core capability to ensure ongoing data quality and observability. Companies can adopt modern tools for data integration, such as highly managed ELT tools like FiveTran or cloud-native ETL platforms like Azure Data Factory. Defining strong governance frameworks early is critical so AI systems can access compliant, trustworthy data from the start.
McElroy draws lessons from successes like American Express and AstraZeneca, which invested in robust data architecture to support continuous learning and feedback loops. American Express built a system for real-time fraud detection, while AstraZeneca uses AI to inform drug discovery and clinical trials within strict compliance boundaries. The lesson is clear: AI success in regulated industries depends on governance as much as innovation.
Trust in AI systems comes from transparency and collaboration between IT and business teams. McElroy recommends a trio of champions—an executive sponsor, business process owner, and technical lead—to ensure alignment across strategy and execution. To balance speed with sustainable infrastructure, he advises resisting short-term wins without a strong foundation and making DRE a core capability. An MIT study shows that repeatable and scalable adoption, not one-off successes, drives sustained ROI from AI.
Governance frameworks should protect the business while accelerating innovation. McElroy suggests creating a governance council with representatives from different business areas and IT. Technical safeguards like tokenizing real data and automating alerts for PII exposure can speed innovation without compromising compliance. To close the skills gap, leaders can create hybrid teams pairing internal staff with external experts, fostering a culture of trust and curiosity around AI. When employees understand how AI supports their work and see its outputs explained clearly, adoption follows naturally.
