Don’t Let AI Reinforce Organizational Silos

In boardrooms and team meetings around the world, the promise of artificial intelligence (AI) looms large. Leaders are told that AI will streamline operations, enhance decision-making and supercharge productivity. But while many organizations are seeing gains from AI, another less visible trend is unfolding—AI is reinforcing functional silos, a problem with which companies have struggled for decades. This can cause performance to go into reverse as departments retreat into their own AI-powered worlds. While each function improves its individual operations, the organization becomes less able to deliver on its corporate strategy.

In this article, we draw on the experience of companies we have researched or advised to look at what happens to performance when AI is adopted in a department-centric way and propose actions to address it (for confidentiality reasons, the names of all the companies referred to below are disguised).

Effect #1: The “technology-first” trap

Very often, department heads decide to implement AI first, and only then look for problems to solve with it. Out-of-the-box AI tools are rarely interoperable, and few teams have incentives to look beyond their own dashboards. What’s more, vendors market these tools as standalone solutions to departments, reinforcing their isolation.

AI tools that remain within department boundaries make it more difficult to address many of today’s most pressing business challenges, such as customer experience, sustainability, and innovation. These require insights and actions that span multiple departments.

Take the case of Emacom (not its real name), an Australian manufacturing company we worked with. Emacom’s IT department implemented AI-powered predictive maintenance, the supply-chain team used separate AI tools for demand forecasting. Meanwhile, the sales department deployed AI for customer service, and HR adopted AI-driven resume screening. Each department operated its AI solution independently.

Rather than addressing the company’s core challenge of reducing delays, these siloed AIs created disconnected efficiency gains but failed to solve problems requiring coordination between maintenance, supply chain, sales, and workforce planning.

How to fix it: Build a “hub and spoke” model

Establish an AI Center of Excellence (CoE) that balances centralized governance with decentralized execution. The CoE acts as the central hub, home to the organization’s top AI experts, strategic leaders, and shared resources. Its purpose is to align all AI initiatives to corporate goals by providing governance, best practices, and a shared infrastructure for scaling AI capabilities. The spokes are embedded AI teams within business functions. They leverage the CoE’s resources and standards to apply their deep domain knowledge, enabling them to solve specific business problems quickly.

Take the case of Bathurst Insurance (not its real name) an Australian company we worked with. Its AI CoE acted as the central hub that identified an opportunity to integrate the company’s sales and underwriting processes. The spokes, embedded teams within sales and underwriting, used the CoE’s shared platform, governance framework and resources to build an AI model that could pre-approve policies in real time and pipe them through to the sales team instantly.

Effect #2: Duplication and contradiction

When different departments use separate data sets and models to solve similar problems, they often reach conflicting conclusions. This isn’t just inefficient; it’s a direct threat to a unified business strategy.

Consider the case of Western Pacific (not its real name), an Australia-based multinational bank we worked with. The finance department’s risk management AI, built on traditional credit scores and historical loan performance, flagged a specific customer segment as high-risk. Simultaneously, the marketing department’s customer acquisition AI, which analysed digital behaviours and social media data, identified the same segment as a prime target for new products. This created significant internal tension. Should the bank aggressively market to these customers, as the marketing team advocated, or should it avoid them, as the finance team advised?

How to fix it: Start with purpose, not process

The fix isn’t to create a universal data set for every team. Instead, it’s about shifting the entire mindset from a “process-first” approach to a “purpose-first” approach. Clearly define the enterprise-wide outcomes you want to achieve, then work backward to determine how AI can support those outcomes across multiple functions. This approach ensures AI is a strategic enabler, not just a tactical tool.

A major Australian online retailer we worked with, which we’ll call Nexora Market, provides a powerful example. Instead of letting departments optimize their individual performance, the company began with a single overarching purpose: to improve customer lifetime value. This led them to create a single, unified recommendation engine. These recommendations became the focus for marketing; inventory management used them to optimize stock levels, logistics used them to predict shipping demands, and customer service used them to offer proactive support.

By focusing on a shared purpose, Nexora Market ensured that all its AI initiatives worked toward a common goal, creating a cohesive and consistent customer experience that transcended departmental boundaries.

Effect #3: Undershot company targets

When AI tools don’t connect and amplify each other, organizations miss the compound effects that make AI truly transformative. Research by one of the authors (Kim) reveals that 70 percent of AI initiatives fail to scale beyond their initial deployment. The reason is precisely because they’re implemented and measured within “siloed AI implementations.”

A major Australian retail chain that participated in Kim’s research, Vera & Wilde (not its real name), celebrated multiple AI success stories across their organization. Their inventory management team reduced stockouts by 15 percent using AI-powered demand forecasting. Customer service improved response times by 40 percent with AI chatbots. The marketing department increased email open rates by 25 per cent through AI-driven personalization.

Each department proudly reported these metrics to leadership and AI adoption was deemed a resounding success. However, despite these individual wins the company’s overall customer satisfaction scores remained flat and competitors were gaining market share. AI was used to optimize departmental metrics rather than to create cross-functional results, such as net promoter score, that would have revealed the brewing problems before they became costly failures.

How to fix it: Incentivize collaboration

Most performance metrics are function specific. Sales chases revenue, HR tracks engagement, operations pursues efficiency. If you want cross-functional AI collaboration, measure it and reward it. Design shared KPIs that reflect collective outcomes like end-to-end customer satisfaction, product launch cycle time, or cross-functional process improvements.

Take, for example, an Australian agricultural field trials company we looked at, CropEdge Research (not its real name). Its departments traditionally operated with siloed KPIs—research teams were evaluated on trial accuracy, sales on client acquisition and operations on cost efficiency. This led to friction: sales promised rapid trial start dates, operations struggled to deliver, and research teams felt rushed. This damaged client satisfaction. To encourage cross-functional AI collaboration, CropEdge Research’s leadership introduced shared metrics, such as client satisfaction scores from trial setup to final reporting; trial turnaround time from contract to delivery; data quality consistency across research, operations, and reporting.

. . .

AI has the potential to unify and elevate your organization, but without deliberate effort, it can just as easily divide it. The fragmentation you’re likely seeing isn’t an inevitable side effect of technology; it’s the result of choices about tool adoption, governance, and culture. So resist the temptation to simply digitize existing silos. Instead, use AI as a catalyst for genuine organizational transformation. Approach AI implementation with the very systems thinking you want your organization to embody. Ultimately, the goal isn’t just to adopt AI, but to use it to build a more cohesive, intelligent, and purpose-driven enterprise.