Delivering ROI Through Autonomous Operations

Gartner recently gathered supply chain, operations and technology executives from across the globe for a Leaders in Action event focused on delivering ROI through autonomous operations.

Special thanks to our Bristol Myers Squibb (BMS) co-hosts Catalina Vargas (SVP, global business unit) and Pádraig Keane (SVP, global supply chain), who welcomed our community to explore BMS's Devens, Massachusetts, campus — a World Economic Forum Lighthouse site that is a microcosm of their global network.

Bristol Myers Squibb shared how they transformed their Product Development & Supply (PDS) organization into a resilient, predictive, and patient-ready digital ecosystem. Their journey highlighted several groundbreaking operational and digital breakthroughs:

• Underpinning them was a set of highly ambitious business goals: halving product development times, billions in cost improvements, and bringing significantly more new products into the pipeline and to market by 2030.
• The PDS organization is moving toward touchless demand planning and will be about 60% of the way there by end of year. Key elements of this new capability include:
  • Live demand signals from orders, market, and macro data.
  • SKU-level ML forecasts that learn from every planning cycle.
  • Scenario planning to stress-test shocks and launches before committing stock.
• Another key innovation highlighted was SmartLaunch, a digital tool designed to simplify and accelerate the launch of new medicines through automated visibility and tollgate governance.
• The team is also leveraging several techniques to unlock higher quality in the data fueling their analytical tools. These include a natural language profiler to improve data management rules, an AI-based data-checking tool, a governed data-rule library, and a data-quality governance reporting platform.
• We toured the Devens facility and walked a path that started in an R&D lab environment, passed through a facility bridging R&D and high-volume manufacturing, and ended with large-scale cell culture manufacturing for mass market products. Having everything on one campus enables faster innovation and improvement cycles.
• Rather than focusing solely on technology, the BMS operations team upskilled over 250 colleagues across more than 100 experiential training sessions and created more than 40 digital-first roles — enabling deployment of more than 30 value-chain use cases.
• To eliminate operational bottlenecks in manufacturing, BMS deployed an electronic batch record platform across clinical and commercial facilities. This solution has expanded yearly batch capacity by more than 20% and reduced time to new product introduction readiness by more than 80%.
• This capability operates alongside AI/ML-driven batch steering models that continuously analyze process data to dynamically optimize production schedules and parameters in real time.

Jim Tobojka, SVP of operations for TE Connectivity, shared his organization’s Integrated Operations Network (ION) transformation, demonstrating how AI, smart machines, and Intelligent Mobile Robots (IMRs) are deployed across the enterprise, including more than a hundred global manufacturing sites.

Tobojka explained how TE cut logistics costs in half by deploying digital control towers, dynamic freight bidding, and multi-modal optimization. While TE has made significant progress in deploying AI agents across its operations, he emphasized that success must be measured by realized business value, adoption, and replication across the enterprise rather than the sheer number of initiatives.

TE Connectivity has a handful of big bet AI applications that are used across all business units (BUs) and it balances this scaled deployment with the ability for individual BUs to innovate their own AI solutions.

There was a lot of discussion in the room about whether supply chain organizations would yield a sufficient return on their AI investments. It is notable that one of the first AI-related solutions that TE Connectivity deployed was a FinOps application that points AI users toward the most cost-optimal solution for their requirements. Sometimes a hammer will do, despite the temptation to use an expensive power tool.

We also had Arun Chandrasekaran, a distinguished VP analyst from Gartner’s CIO practice, share perspectives with our community. One concept that he used which was newer to some leaders was that of “data gravity.”

Data gravity refers to the way large datasets attract applications and services to where the data is stored. Chandrasekaran described the need to shift toward “semantic gravity,” which is how meaningful, well-interpreted information draws AI and agentic initiatives that depend on that meaning.

Chandrasekaran also highlighted that we’ve seen a more than 90% decrease in cost per token over the past year. Unfortunately, with proliferating agents running recursive calls, token consumption has increased up to 30 times in some companies leading to a much higher Total Cost of Ownership (TCO) for AI solutions.

In response, leading organizations are:

• Starting to measure cost per successful outcome to avoid "cheap token traps."
• Aggressively targeting wasteful context and undisciplined AI prompting.
• Establishing hard user token budgets, session ceilings, and automated kill switches to prevent runaway recursive loops.

We are grateful for the inspiring insights shared during our time together at BMS in Massachusetts and look forward to continuing these strategic discussions across our COO and CSCO community.

Stan Aronow
VP Distinguished Advisor
Gartner Supply Chain
Stan.Aronow@gartner.com