AI Is Not Ozempic for Supply Chains
Wall Street wants AI to be a miracle drug for operations. But the COOs on the hook to deliver the results say it’s more like a demanding workout regimen.
CEOs are under intense pressure from investors to quickly harvest the productivity benefits of AI and agentic systems in operations. But there’s a problem: their operations leaders aren’t so sure. In a recent Zero100 survey of 100 COOs – the executives responsible for delivering on these AI promises – we uncovered a stark disconnect between CEO commitments to investors and the reality on the ground.
Most say that less than 50% of what has been promised is possible within stated timeframes. Forty-two percent believe that less than a quarter of their CEO’s AI-related commitments are realistic.
Wall Street wants to believe AI is like a miracle weight loss GLP-1 drug. Operations leaders see it more like a new exercise program that can turn anyone into a better athlete – transformational with targeted effort, but no magic pill.
Escaping Pilot Purgatory
We recently hosted a roundtable comprising 13 COOs from retail, pharmaceutical, industrial, CPG, and consumer electronics companies. None doubted the power of AI to enable dramatic improvements, but all agreed that escaping “pilot purgatory” requires a level of process discipline not often discussed on CNBC. In other words, investor AI hype, while not necessarily wrong, underestimates the foundational work and focus needed to scale AI beyond test-and-learn.
Focus comes first. One of our panelists, for example, specified four targeted areas to address with AI in 2026. Another identified eight across operations, engineering, and commercial workflows. Many see clear paths to payback in forecasting, planning and procurement workflows. Operations leaders are past the star-struck phase of discovering generative AI as a personal productivity tool and are well into imagining how agents can be a force multiplier for productivity. They now need to get specific about deployment plans.
Foundational work comes second. Our panelists all talked about good data as a prerequisite for success with AI, but many felt confident about their progress in developing their ontologies, data governance processes, and technology infrastructure. Also cited as important foundational work is skill development around AI tools, plus the human-to-human collaboration required to manage agents as partners in human-machine processes. Last, but certainly not least, is ensuring decent user support for tools that are deceptively easy to use, but don’t necessarily send error messages when they’re wrong.
Workflow Thinking Is the Unlock
One of our panelists crystallized the main challenge (and opportunity) facing COOs now: “The big game changer is to rethink and redesign the workflow where the human and the non-human should be focused on impact.” Going back to our exercise analogy, workflow thinking is like having an expert personal trainer targeting specific muscle groups with a structured program – strategic, focused, and designed for sustainable results.
Data from our COO survey shows that most operations leaders already think this way. A majority expects no more than 25% of current operational workflows to be rebuilt or redesigned by agentic AI in the next two years. Overall, nearly one in five plan to redesign less than 10% of all workflows in that time. They are not trying to reinvent everything – just those workflows with a clear path to results.
Process Discipline Is the COO’s Secret Weapon
Another panelist cited as a benchmark financial services’ relative success with AI, noting that as a highly regulated industry, their processes are standardized, and their data is relatively clean. This nugget is telling because process discipline not only drives quality and safety in manufacturing but also helps keep data consistent through time. It is also the core of much operations gospel from Six Sigma to the Toyota Production System.
The top four process flows Zero100 members are working on with AI and agentic systems now include:
- Inventory replenishment and optimization
- Source-to-contract and plan
- Barrier-free demand-to-product
- Manufacturing excellence & automation
Each of these has big potential payback, data requirements that don’t require boiling the ocean, and embedded tasks that will benefit from generative or classical AI. They are mission-critical processes cutting across functions to drive enterprise-level results. This means intense process discipline is essential to avoid siloed suboptimization.
COOs are uniquely positioned to define, measure, and reinvent these processes.
Fit to Win
AI is great, but it is no miracle cure for badly run operations. Far better to focus on the process outcomes you can commit to and do the foundational work on data, people, and support.