Share with your CIO
Most enterprise AI programs are failing not because the models are weak but because the organizations running them aren’t ready, according to this analysis drawing on Jaclyn Rice Nelson’s perspective, CEO of Tribe and former VP at Google Capital. Nelson’s argument centers on three compounding failure points: vague problem definitions, absent change management, and incentive structures that punish the cross-functional behavior AI integration actually requires. The piece extends the argument to emerging markets, citing a hypothetical KES 150 million AI rollout in Kenya as a case where missing data governance destroys the investment entirely.
What this means for your business
The organizations most exposed here aren’t the ones that haven’t started their AI programs. They’re the ones that have. A CIO who has already committed budget to a generative AI deployment is now sitting on a system whose output quality is a direct reflection of data plumbing decisions made years ago, by teams that weren’t thinking about AI at all. If your data is fragmented across business units, the model doesn’t fix that. It amplifies it, producing confident-sounding errors that erode trust faster than any failed pilot ever could.
Nelson’s core claim, shaped by her position advising companies that need to believe AI adoption is solvable with the right consulting framework, still holds on the merits. The change management point is understated in most enterprise AI coverage, which tends to treat workforce resistance as a communication problem rather than a rational response to poorly designed deployment. When employees encounter an AI tool that lacks context for their actual work, reverting to the spreadsheet isn’t stubbornness. It’s efficiency. The implication for CIOs is that the sequencing of an AI program matters as much as the selection of tools, and most programs get the sequence wrong by leading with the interface rather than the infrastructure.
The incentive structure argument is the sharpest and least actioned insight in the piece. Most enterprises still reward individual and departmental performance, which creates a structural disincentive for the data sharing that AI systems require to function well. A sales org that hoards its CRM data to protect its metrics is making a locally rational choice that globally degrades the AI model. CIOs who can’t point to explicit changes in how cross-functional AI collaboration is measured and rewarded should treat their current program’s ROI projections with real skepticism. The budget renewal conversation changes shape when that gap is visible.
Concept deep-dive: Data governance
Data governance is the set of policies and processes that determine who owns enterprise data, how clean it must be, and who can access it for what purpose. It exists because large organizations accumulate data across dozens of systems that were never designed to talk to each other. Think of it as the difference between a library with a card catalog and one where books are piled randomly in rooms. AI models trained on ungoverned data inherit every inconsistency, and the outputs reflect that chaos.
Based on reporting from Delivering Enterprise AI Value Requires More Than Technology, originally published 2026-07-12 03:00:00.
