On September 18, 2025, Harvard Business Review published 'Don’t Let AI Reinforce Organizational Silos.' The article by Graham Kenny and Kim Oosthuizen warns that while AI is driving productivity inside teams, it also reinforces silos—where departments optimize themselves at the expense of company-wide coordination. The piece highlights real-world cases of marketing copilots, finance forecasting models, and HR automation running on separate pipelines and calls for leadership-level interventions.

AI isn’t just another tool—it changes work patterns, decision velocity, and information flows. When teams adopt AI in isolation, they often optimize for local KPIs, lock data into inaccessible models, and create duplicated effort. This risks faster but less coordinated decisions, fragile integrations, and missed opportunities to scale AI’s value across the enterprise.

The article was authored by Graham Kenny (Strategic Factors) and Kim Oosthuizen (Head of AI, Bupa Australia & NZ). Their concerns are backed by research from McKinsey, Gartner, and Berkeley’s California Management Review, which all point to similar patterns: local AI successes but enterprise-wide integration struggles.

What are the key takeaways?

AI can amplify existing silos if deployed in isolation. Data integration and governance determine whether AI unites or fragments organizations. A centralized or federated approach to standards, APIs, and measurement helps scale AI beyond pilots.

What are the wider implications?

If leaders ignore silo effects, AI may become a set of competing micro-optimizations that harm strategic outcomes. Financially, this leads to duplicated costs and failed scale-ups. Operationally, it creates fragile integrations and inconsistent outputs. Strategically, it misaligns incentives—teams chase local KPIs at the expense of customer experience and corporate goals.

What should leaders do?

Leaders should create a federated AI governance model, invest in shared data platforms and APIs, align metrics to corporate outcomes (not just departmental KPIs), and treat AI projects as long-term products with monitoring, maintenance, and cross-team SLAs built in.

AI adoption challenges in Reinforce of Silos -

AI adoption challenges, Harvard Business Review authors Graham Kenny and Kim Oosthuizen warn that, while AI is driving productivity inside teams, it’s also quietly hardening organizational silos—letting departments optimize themselves at the expense of company-wide strategy and coordination. Their piece argues leaders should treat AI as a cross-functional design problem, not just a set of departmental tools. (Harvard Business Review)

What happened? (facts, timeline, announcement details)

On September 18, 2025, HBR published “Don’t Let AI Reinforce Organizational Silos,” a clear call to action for executives deploying AI across functions. The article documents how pockets of AI adoption—marketing copilots, finance forecasting models, HR automation—deliver local wins but often run on separate data pipelines, goals, and governance, which compounds long-standing coordination failures rather than solving them. The authors outline how this pattern looks in practice and propose leadership-level interventions. (Harvard Business Review)

Why it matters

AI isn’t just another tool—it changes work patterns, information flows, and decision velocity. When AI systems are built and measured by isolated teams, they can:

Optimize for a department KPI rather than the corporate objective.
Lock data into proprietary models that other teams can’t access.
Create duplicated effort and fragile integrations that are expensive to undo.

That means short-term efficiency gains can translate into long-term strategic drift: faster but less coordinated decisions, brittle operations, and missed opportunities to scale AI’s value across the business. The concern is supported by broader AI adoption research showing rapid experimentations inside teams but slow organizational integration. (McKinsey & Company)

Who’s involved

Authors & experts: Graham Kenny (Strategic Factors) and Kim Oosthuizen (Head of AI, Australia & NZ, Bupa) wrote the HBR piece and offer practical leadership advice. (Harvard Business Review)
Research & consultancies: McKinsey and Gartner studies show the same pattern—heavy experimentation and pilot activity in business units, but scaling and integration are hard and often fail without cross-functional governance. (McKinsey & Company)
Academia & practitioners: Recent commentary (e.g., Berkeley’s California Management Review) frames the issue as “the silo effect in the AI age,” echoing HBR’s prescription to combine technical and organizational fixes. (California Management Review)

Key takeaways (bullet quick-read)

AI can amplify existing coordination problems if deployed in isolation. (Harvard Business Review)
Data integration and governance are the levers that decide whether AI unites teams or fragments them. (CMSWire)
A centralized—or at least federated—approach to standards, APIs, and measurement helps scale AI beyond pilots. (McKinsey & Company)

Expert perspective (quotes & interpretation)

“AI will streamline operations—but without intentional design it will accelerate the walls between teams,” say HBR authors Graham Kenny and Kim Oosthuizen. Their point: leadership must reframe AI rollout from a series of local automations into a company-level systems design problem. (Harvard Business Review)

Analysts at McKinsey and Gartner back this up: many organizations report strong ROI in pockets, yet struggle to integrate models into enterprise workflows or to maintain them at scale—some agentic AI projects are even being scrapped when the integration and governance burdens surface. (McKinsey & Company)

Wider context — how this fits current AI trends

Experimentation is rampant: Employees and individual teams are adopting generative AI and task automation fast; organizational maturity lags behind. (McKinsey & Company)
Data is the bottleneck: Many AI failures aren’t models but fractured data pipelines and ownership disputes. Fixing data governance gives the biggest multiplier for AI in production. (CMSWire)
Governance & ops matter: The shift from prototyping to production exposes issues of tooling, monitoring, security, and cross-team SLAs—areas where siloed projects fail first. (McKinsey & Company)

Short expert analysis — federated AI model

If leaders ignore this problem, organizations risk turning AI into a set of competing micro-optimizations that slow down strategic outcomes. The practical implications:

Financial: Duplicate engineering and data costs, wasted licensing fees, and failed scaleups of point solutions. (Reuters)
Operational: Fragile integrations and inconsistent model outputs that confuse downstream processes. (CMSWire)
Strategic: Misaligned incentives where teams optimize for local KPIs at the expense of customer experience or company goals. (Harvard Business Review)

Concrete actions for leaders (short list):

Create a federated AI governance model: central standards + local ownership. (McKinsey & Company)
Invest in shared data platforms and APIs so models can interoperate. (CMSWire)
Align measurement to corporate outcomes, not only departmental KPIs. (Harvard Business Review)
Treat AI projects as long-term product bets with maintenance, monitoring, and cross-team SLAs built in. (Reuters)

Let’s talk!