Why Most CPG AI Pilots Stall Before They Scale

An Industry Betting Big on a New Technology

CPG leaders aren’t waiting on the sidelines. The IBV research found that more than two-thirds of industry leaders agree generative AI is important to their organization’s future, and more than three-quarters believe it should be adopted quickly to keep pace with competitors.

Adoption is already underway in specific functions. Nearly half of CPG organizations are piloting or implementing generative AI in supply chain, logistics, and fulfillment, and 40% are doing the same in product development.

But there’s a telling gap in that data: only 3% of organizations report they’ve reached the optimizing stage in either function. 

In other words, most companies are still early in translating pilots into scaled, reliable value, and that pattern shows up across nearly every AI initiative that skips foundational data work. That same gap points to where the real opportunity (and the real risk) lives.

The Use Cases Already Paying Off

The Institute for Business Value’s findings highlight specific, ranked use cases where CPG and supply chain leaders see the most value today. Those use cases can be broken up into product development and supply chain contexts:

In product development:

• Augmenting internal and external knowledge search (97%)
• Creating analytical models (92%)
• Generating custom product visualization and designs (91%)

In supply chain:

• Augmenting knowledge search (95%)
• Generating demand forecasts from structured and unstructured data (89%)
• Summarizing complex documents (81%)

Demand forecasting stands out as a particularly high-value application. IBV’s benchmark survey found that nearly half of executives expect generative AI to reduce forecast error by roughly 20% and cut inventory carrying costs by close to a quarter. Those numbers translate directly into working capital freed up and waste avoided.

The report also points to less obvious applications with outsized impact: using generative AI to model digital twins of the supply chain before a product launches, reducing costly late-stage product changes, and improving how trade promotions and sourcing decisions get made. Across all of these, the common denominator is the same: generative AI works best when it’s synthesizing large, messy, often unstructured pools of information into something a decision-maker can actually act on.

When Sustainability Becomes a Data Problem

One of the more forward-looking findings in the report is how tightly sustainability and regulatory compliance are becoming linked to AI strategy. Most of a product’s lifetime environmental footprint and cost structure is locked in at the design stage. In other words, the earlier sustainability data can inform decisions, the more impact it has.

The numbers back this up: 69% of consumer product executives say generative AI will be important to their sustainability agenda, and 73% plan to increase investment in generative AI specifically for sustainability use cases. Meanwhile, 84% agree that high-quality, transparent data is a prerequisite for hitting sustainability goals at all.

Regulatory compliance shows a similar pattern. One global consumer products company referenced in the report built a generative AI assistant to manage more than 1,000 regulations across its worldwide operations—cutting analysis time from hours or days down to minutes, with a projected profit impact north of $165 million over five years.

The Bottleneck Hiding Behind Every Pilot

Perhaps the most important finding in the report is also the least flashy: when IBV asked CPG leaders about the top six barriers to generative AI adoption, half of them were data-related. Concerns about data accuracy and bias (52%), limited access to underlying technology (46%), and insufficient proprietary data to customize models (40%) were some of the most noteworthy.

Just over half of consumer CEOs say a lack of clear standards is delaying investment altogether, and nearly two-thirds admit they lack consistent standards for data, privacy, or sustainability in at least one strategic area.

This is the crux of it: generative AI is not a plug-and-play solution, and it isn’t a one-size-fits-all replacement for traditional automation or machine learning. The organizations seeing real returns are the ones treating data readiness as the prerequisite, not an afterthought.

Turning Readiness Into a Real Strategy

The report’s action guide boils down to a few practical priorities that echo what we see across the industry every day:

• Build a trustworthy data foundation first. Know which of your data is structured and reliable enough for traditional automation or ML, and which unstructured data is better suited to generative AI.
• Start with high-value, well-scoped use cases rather than broad, undirected experimentation — and be honest about which problems generative AI is actually the right tool for.
• Bring suppliers and partners into the data strategy early. Some of the report’s most compelling examples, like fragrance maker Symrise unlocking two million historical formulas with an AI assistant, or British Sugar cutting knowledge-retrieval time from 20 minutes to 20 seconds, only worked because the underlying data was accessible and well-organized in the first place.
• Treat sustainability and compliance as data problems, not just reporting obligations — the earlier they’re embedded in decision-making, the more value they create.