Most enterprise AI pilots never become systems the organisation actually depends on. According to Bharat Chadha, Partner for Tech Consulting at Uniqus Consultech, the single biggest bottleneck is the data and context layer, not the model. Pilots succeed on curated datasets built for demos. Production requires real-time access across ERP, CRM, documents and workflows while respecting permissions, business rules and regulations. Once users cannot trust the source, quality or security of the data, they stop trusting the output. Models themselves are becoming commoditized; the durable advantage is a trusted data foundation. Cost predictability is the second barrier. AI spend scales with usage, and that variability stalls the approvals needed to move pilots into core systems. Chadha recommends a barbell talent approach: keep core technology teams modernizing the estate while pushing frontier skills into the business so domain experts can build agents inside central guardrails.
Most enterprise AI pilots never become systems the organisation actually depends on. According to Bharat Chadha, Partner for Tech Consulting at Uniqus Consultech, the single biggest bottleneck is the data and context layer, not the model. Pilots succeed on curated datasets built for demos. Production requires real-time access across ERP, CRM, documents and workflows while respecting permissions, business rules and regulations. Once users cannot trust the source, quality or security of the data, they stop trusting the output. Models themselves are becoming commoditized; the durable advantage is a trusted data foundation. Cost predictability is the second barrier. AI spend scales with usage, and that variability stalls the approvals needed to move pilots into core systems. Chadha recommends a barbell talent approach: keep core technology teams modernizing the estate while pushing frontier skills into the business so domain experts can build agents inside central guardrails.