Enterprise AI Mandates Face Organizational Barriers

Deploying enterprise-grade AI across a large organization requires managing complex data estates, securing compute availability and making systems work together. But Prashant Prahlad, Capital One Software’s SVP of product, told CIO Dive that technical issues are often among the least of enterprises’ challenges. He leads product management for its enterprise B2B software business, which helps customers prepare data to scale AI, transform data and migrate to the cloud. He said he often advises customers on building organizational consensus and momentum around AI.

Prahlad said organizations struggle to rally around AI when teams have different incentives. Many have a mandate to adopt AI first but lack the authority to align people across the organization. Employees are more likely than senior staff to distrust AI and report greater personal risk from autonomous systems because they have less control. Prahlad said a strong data foundation underpins AI scaling, while project failures have risen in recent years because of poor data readiness rather than technology limits.

Measuring return on investment is difficult when the foundations for an accurate assessment are missing, Prahlad said. Most businesses have not justified their AI spending and say they need to put systems into production and provide models with proprietary sensitive data to test them properly. That is difficult while CIOs must report value to boards; most organizations are still not seeing cost efficiency or productivity gains to the same extent that AI is helping generate insights or support decisions.

Prahlad also said vendor promises are flooding organizations, and existing products that can be prepared for AI may be a better choice. AI projects emerging across departments can make it hard for central teams to distinguish useful signals from noise. He recommended starting with one use case and forming a cross-functional “tiger team” to identify and test data. An IT team could deploy a chatbot in one part of the organization, then examine the data readiness, governance and security needs.

One client used a single use case to assess its data and found it lacked a centralized data strategy. It then created a team to centralize data management across seven business lines, Prahlad said; the pilot helped identify organizational gaps. He said readiness varies widely, and large technology teams do not guarantee that an organization is ahead. Progress depends on how deliberately it adopts AI and the mandate it has received, he said.

Deploying enterprise-grade AI across a large organization requires managing complex data estates, securing compute availability and making systems work together. But Prashant Prahlad, Capital One Software’s SVP of product, told CIO Dive that technical issues are often among the least of enterprises’ challenges. He leads product management for its enterprise B2B software business, which helps customers prepare data to scale AI, transform data and migrate to the cloud. He said he often advises customers on building organizational consensus and momentum around AI.

Prahlad said organizations struggle to rally around AI when teams have different incentives. Many have a mandate to adopt AI first but lack the authority to align people across the organization. Employees are more likely than senior staff to distrust AI and report greater personal risk from autonomous systems because they have less control. Prahlad said a strong data foundation underpins AI scaling, while project failures have risen in recent years because of poor data readiness rather than technology limits.

Measuring return on investment is difficult when the foundations for an accurate assessment are missing, Prahlad said. Most businesses have not justified their AI spending and say they need to put systems into production and provide models with proprietary sensitive data to test them properly. That is difficult while CIOs must report value to boards; most organizations are still not seeing cost efficiency or productivity gains to the same extent that AI is helping generate insights or support decisions.

Prahlad also said vendor promises are flooding organizations, and existing products that can be prepared for AI may be a better choice. AI projects emerging across departments can make it hard for central teams to distinguish useful signals from noise. He recommended starting with one use case and forming a cross-functional “tiger team” to identify and test data. An IT team could deploy a chatbot in one part of the organization, then examine the data readiness, governance and security needs.

One client used a single use case to assess its data and found it lacked a centralized data strategy. It then created a team to centralize data management across seven business lines, Prahlad said; the pilot helped identify organizational gaps. He said readiness varies widely, and large technology teams do not guarantee that an organization is ahead. Progress depends on how deliberately it adopts AI and the mandate it has received, he said.