Pharmaceutical companies risk wasting significant resources by rushing into artificial intelligence deployments without first establishing a coherent agent strategy, according to senior Cencora leaders.
Matt Sample and Gregg Gorniak, both executives at healthcare distribution giant Cencora, have outlined a disciplined framework for how the industry should approach the rapid proliferation of AI agents.
Their central argument is one of sequencing: organisations that chase attractive individual use cases before defining a broader strategic architecture are likely to create fragmented, hard-to-govern AI environments.
The warning carries weight given Cencora’s scale and its deep operational involvement across pharmaceutical supply chains, where decisions carry direct consequences for patient outcomes.
At the heart of their position is a firm philosophical boundary around what AI agents should and should not be permitted to do within high-stakes healthcare settings.
Sample and Gorniak argue that AI in pharma should occupy an advisory role, surfacing recommendations and analysis for human review rather than autonomously executing consequential decisions.
This “recommend, not decide” principle reflects a broader tension playing out across regulated industries, where the speed of AI capability development is outpacing governance frameworks.
The executives suggest that pharma’s unique regulatory environment and patient safety obligations make the case for human oversight particularly compelling compared with other sectors embracing autonomous AI action.
By insisting on human decision authority, companies can capture the efficiency benefits of AI-driven analysis while maintaining the accountability structures that regulators and patients expect.
The Cencora position also implicitly pushes back against vendor narratives that frame agentic AI as a straightforward productivity upgrade ready for immediate, broad deployment.
Instead, Sample and Gorniak advocate for organisations to define what an AI agent is permitted to act on, what it must escalate, and how those boundaries are monitored before any use case goes live.
This architecture-first thinking mirrors approaches that have emerged in financial services, where firms learned through costly experience that ungoverned automation can amplify errors at scale.
For pharma supply chains specifically, where stock availability, cold chain integrity, and distribution timing intersect with clinical need, the consequences of miscalibrated AI action could extend well beyond operational inconvenience.
The Cencora executives appear to be positioning the company as a thoughtful voice in an industry conversation that has, in many quarters, been dominated by enthusiasm over rigour.
Whether peers across the pharmaceutical distribution and manufacturing landscape heed that call for strategic patience before deployment will likely shape how smoothly the sector navigates its AI transition.