CSCOs embrace a multiagent future: talent, ROI, and governance

Bridging the talent gap and retaining functional expertise

As supply chain organizations accelerate their adoption of AI, the most complex challenge isn't the technology itself, it is the human element. The true bottleneck to scaling AI effectively lies in acquiring, developing, and deploying the right talent. To navigate this transformation, supply chain leaders are increasingly adopting a strategic, hybrid approach to workforce development.

For highly technical roles, such as data scientists and data engineers, organizations are building from the ground up, often recruiting from universities. This allows early-career professionals to gain broad, foundational exposure across various business functions, embedding their technical capability deep within the enterprise.

When it comes to functional and leadership roles, companies are looking externally to acquire experienced transformation individuals that one CSCO referred to as athletes. Bringing in external leaders who possess mature best practices and fresh perspectives provides the necessary momentum to push the envelope on innovation, working in tandem with established procurement and supply chain leaders.

In distribution-focused organizations, this approach helps accelerate new ideas and ways of working. As described by a CSCO of a distribution company: “We brought in a procurement transformation leader who sits alongside the procurement leader. It’s pushed the envelope on new ideas.”

Integrating advanced AI tools into daily operations introduces a significant strategic risk: the potential deskilling of the workforce. If teams begin accepting AI-generated output without fully grasping underlying business nuances, the organization could lose its critical subject matter expertise. Foundational knowledge cannot be outsourced to an algorithm.

According to a technology CSCO, there is growing concern that teams may rely too heavily on generative outputs, losing sight of what many describe as a core principle of AI adoption: “human first, human last.”

The future of supply chain talent strategy must emphasize the enduring requirement of the "human in the loop." AI must be positioned to augment, rather than replace, human decision-making, particularly in high-stakes areas such as strategic negotiations, relationship management, and complex exception handling.

The most valuable skills moving forward will not be purely technical, but rather adaptive and cross-functional. Employees must be trained not just to use AI, but to orchestrate human-machine collaboration effectively, ensuring that a human in the loop principle governs all AI-driven processes. By prioritizing the workforce, organizations can unlock the AI without sacrificing the functional expertise required to run a resilient supply chain.

“How do we set up the org structure, so agents and humans are working together?”

CSCO for a technology company

Moving from experimentation to execution

Supply chain leaders are moving past AI experimentation and focusing on execution. Across the industry, the mandate is clear: AI initiatives must scale and deliver value, as well as measurable return on investment (ROI). While there is no universal library of AI use cases for supply chain, practical applications are emerging across advanced planning, procurement, and risk management.

A primary focus is leveraging AI to transform traditional forecasting into dynamic, predictive modeling. “Many supply chain leaders are focused on providing better insights into their customers upstream and downstream to better determine the reason for shortages,” said a CSCO for a pharmaceutical company. “We built out an insights dashboard to be more predictive in analyzing market events.”

Beyond demand planning, organizations are deploying AI agents to tackle labor-intensive, operational bottlenecks. For instance, in one case an internal AI agent has been developed to automatically roll up material shortages across the entire company on a weekly basis. This agent doesn’t just aggregate data; it synthesizes information, suggests actionable next steps for procurement teams, and generates emerging trend reports for leadership, thus automating a previous manual process. Additional practical applications involve constructing predictive dashboards for weather forecasting, monitoring geopolitical disruptions, and employing AI to dynamically evaluate Harmonized Tariff Schedule (HTS) risks in real time.

The journey from experimentation to execution is not without its challenges. The early assumption that advanced generative AI could magically make sense of completely unstructured or messy data has proven overly optimistic. 

Chris McCarney, KPMG Procurement and Supply Chain leader, points to how early hype around AI’s ability to work with unstructured data is being recalibrated. “18 or 24 months ago, we would have said, ‘don’t worry about structured data.’ Now the better data is, the better it will work and perform.”

Leaders agree that the quality of underlying enterprise data, especially within core ERP systems, remains the fundamental prerequisite for AI success. You cannot build a predictive, autonomous supply chain on top of fragmented data. Ultimately, realizing the full ROI of AI programs requires as much rigorous investment in data management and governance as it does in the algorithms themselves.

"We built an AI agent to automate our reporting process. What used to be a manual burden is now automated, providing actionable insights to our leadership." CSCO for a technology company.

Balancing grassroots AI innovation with centralized governance

As AI tools become democratized across the enterprise, leaders are confronting a new, unintended consequence: AI sprawl. The core challenge is no longer just how to build an AI agent, but how to govern a complex, multi-agent environment. The ease with which different business units spin up localized agents is rapidly creating a fragmented landscape.

Enterprise software providers, from legacy ERPs to modern procurement suites, are encouraging companies to build AI capabilities on their platforms. Organizations risk creating a tangled web of disconnected algorithms. As many supply chain leaders describe the challenge, “we’re dealing with a proliferation of people creating their own agents, doing their own thing, but with no orchestration or understanding of what's going on in one silo versus another.”

While agent proliferation is the challenge, there are also examples of employees using agents to solve real operational challenges. In one instance, a supply chain leader described how, “It rolls up every week our shortages across the company. That would take at least a person and a half. The process is now automated.”

These standout cases are rare. The uncoordinated growth of agents more often leads to duplicated efforts, conflicting data interpretations, and a severe lack of scalability. To combat this, leaders are exploring creative ways to map out their hybrid workforce. One emerging best practice calls for placing AI agents on organizational charts alongside human employees, complete with brief job descriptions. This visual management approach helps leadership identify where algorithmic efforts are being duplicated and clarifies the division of labor across the supply chain.

However, visualization is only the first step; true orchestration requires robust, centralized governance. The conversation among CSCOs is rapidly shifting toward the necessity of an agent control tower, just as a traditional supply chain control tower provides visibility into physical inventory and logistics.

Strategic oversight is essential to manage agent-to-agent interactions, ensure interoperability across existing technology stacks, and establish critical guardrails—including kill switches to rapidly shut down agents that drift from their intended purpose or execute unauthorized actions. Ultimately, the organizations that dominate the next decade of supply chain innovation will not be those with the highest volume of independent AI agents, but those with the most cohesive strategy for orchestrating them.

"We are going to need agent control towers with governance and kill switches in place if an agent isn't working the way we want it to."

Chris McCarney, KPMG Procurement and Supply Chain leader