From pricing to personalization, AI is moving closer to the consumer, exposing whether companies can scale with visibility, trust and control.
By Aili McConnon, News Writer | Inbound Marketing Editorial Strategist, IBM Think
The executive brief
- AI governance has shifted from a risk control function to a performance lever that’s linked to faster scale, lower spend and stronger margins.
- The constraint isn’t launching AI; it’s keeping control once systems start acting across the business, which is why governance-by-design is essential.
- In consumer-facing industries, weak governance becomes brand risk, where a single bad output can erode trust in real time.
- Agentic AI raises the stakes by exposing how quickly experimentation is outpacing the visibility, accountability and cost control needed to run it.
- Leading organizations treat governance like infrastructure: embedded early, reused often and built to scale with speed and control.
The problem with scaling AI
As consumer goods companies race to embed AI into demand forecasting, pricing, marketing, customer service and product innovation, many are discovering that the issue isn’t whether they can easily launch another pilot. It’s whether they can keep control once AI starts making, recommending or automating decisions across the business. Enter responsible AI.
A growing body of evidence suggests governance is what allows AI to scale. According to new IBM Institute for Business Value (IBV) research, organizations that embed control into their AI architectures deploy 16 times more AI agents, spend four times less and achieve 18% higher operating margins than peers that do not. The finding reframes responsible AI from a defensive function to manage risk into a core performance lever. And governance will only become increasingly important as Gartner, for example, predicts that by 2027, AI governance and responsible AI capabilities will be part of 75% of AI platforms, making them the main area of competition.
The false choice between speed and control
For CEOs, CIOs, CTOs and chief AI officers, the governance debate is shifting. Many organizations still treat governance as something that can follow adoption: launch the pilots, prove the use case, add controls later. Others respond to risk by slowing everything down in the name of safety. Neither approach scales AI quickly and safely throughout an organization.
“Everyone needs a driver’s license—AI needs one, too. We are not there yet.”
Rudy Hagedorn
Director of the data driven value chain
CEO-led Consumer Goods Forum
The companies pulling ahead are trying to avoid that tradeoff by building rules into how AI systems work from the start, including what data an agent can access, what actions it can take, what evidence it must produce and when a human must intervene. In that model, governance isn’t a committee that reviews AI after the fact; it’s a system of trust designed into data pipelines, models, workflows and decision points.
As Rudy Hagedorn, director of the data-driven value chain at the CEO-led Consumer Goods Forum, put it in an interview with IBM Think, the gap today is similar to letting somebody operate powerful machinery without certification: “Everyone needs a driver’s license—AI needs one, too. We are not there yet.”
When AI becomes a brand risk
Closing the gap is especially urgent in consumer packaged goods (CPG), where AI is rapidly moving into areas that shape consumer trust: pricing, claims, personalization, promotions, supply forecasting and brand communications. One wrong recommendation or unchecked output can become a brand incident. A poorly monitored AI system can generate incorrect pricing, misleading claims or biased personalization right at the moment when a consumer is making a decision.
Phaedra Boinodiris, Global Leader for Trustworthy AI at IBM Consulting®, told IBM Think that many organizations still lack the basics needed to manage that risk. “You can’t possibly govern what you can’t see,” she said, noting that companies need visibility into their AI inventories, audits, regulatory exposure and model behavior. Governance, she added, can’t stop at legal compliance because “you can have AI be lawful but awful.”
Agentic AI raises the stakes
As companies move from generative AI pilots to agentic AI systems that can act more autonomously, the risk if something goes awry multiplies manifold. Boinodiris said IBM Research found that while 88% of organizations are experimenting with agentic AI, 80% still lack foundational governance for earlier forms of AI. That gap, she said, makes literacy, accountability and authority essential. “The work of governance,” she said, “requires power, and it requires a funded mandate. It’s not a side gig.”
As AI agents begin acting on behalf of employees, customers or business functions, control can no longer sit outside the system. It has to be a part of how the system operates. Leaders need to know not only what an AI tool is doing but whether it’s acting within approved boundaries in real time.
The hidden cost of ungoverned AI
The issue of governance isn’t only reputational. It’s financial. As AI usage spreads across business units, models, vendors, infrastructure and tokens can quickly become a new form of shadow spend. The IBV research notes that 84% of tech leaders haven’t fully operationalized AI financial management, while 85% lack full visibility into real-time AI spend. Without accountability, AI costs can rise even when business value remains unclear.
At the same time, many companies are overlooking a foundational constraint: data quality, Hagedorn said. As he put it, “AI feeds on data—look at what it eats.” What many organizations have is “a mound of data—not really interoperable,” often fragmented across decades of systems and silos.
84% of tech leaders have not fully operationalized AI financial management
85% lack full visibility into real-time AI spend
That creates a familiar risk, with higher stakes. “Garbage in, garbage out” is no longer a back-office problem—it directly shapes AI outputs at scale. Increasingly, companies are exploring the use of AI itself to clean, standardize and validate data before it’s ingested into models, recognizing that “responsible AI use is mostly about the data use,” Hagedorn said.
That’s why governance is increasingly tied to scale. AI programs often stall, not because leaders lack ambition, but because they lack repeatable controls, Boinodiris said. If each use case requires a bespoke approval process, bespoke architecture and bespoke risk review, AI stays stuck in pilots. But if the same control rules are embedded once and reused across systems, each new use case can move faster within known boundaries.
From pilots to repeatable performance
Elaine Parr, Vice President of Consumer Industries at IBM Consulting, said the organizations making progress are those that think beyond isolated proofs of concept. “It’s not what you’re AI-ifying; it’s how you’re doing it,” she said. “Unless you’re thinking revolution, unless you’re thinking about those higher-order benefits, you’re always going to be in this kind of iterative POC frenzy.”
“Unless you’re thinking revolution, unless you’re thinking about those higher-order benefits, you’re always going to be in this kind of iterative POC frenzy.”
Elaine Parr
Vice President of Consumer Industries
IBM Consulting
Parr pointed to the resurgence of process ownership as one way large companies are trying to govern AI transformation. The old idea of a global process owner is becoming, in her words, more like “global activity ownership”—someone who oversees both the process and the platform, data, skills and operating model that surround it.
Treat AI risk like supply chain risk
That operating discipline is familiar territory for CPG leaders. These companies already manage product recalls, supplier risk, quality assurance and regulatory audits. The next step is applying the same rigor to AI, with incident tracking, severity classification, real-time intervention protocols and clear ownership when systems fail.
Some industry leaders are already moving in that direction. Parr pointed to Nestlé as a CPG company that has been “very thoughtful and intentional” about its transformation. She noted that Nestlé has publicly discussed its Nestle Fuel for Growth transformation strategy, which is targeting CHF 3 billion in benefits by the end of 2027. Parr framed the initiative as not simply a cost program but a governance story about sequencing transformation, workforce change and technology decisions.
Governance by design, not by checklist
In the AI work IBM does with Heineken, the team treats responsible AI as an enabler, not a blocker, said Sophie Kuijt, IBM Distinguished Engineer, to IBM Think. Speaking about the approach applied with Heineken, she said governance is a priority “by design” as it scales new technologies and use cases. The goal, she said, is to make clear from the beginning that innovation must meet defined principles. Governance is “not used as a blocker” or merely a compliance checklist but as “an enabler” that builds transparency early in the process.
This shift mirrors a broader operational reality: governance can’t remain an external layer. As Hagedorn described it, companies increasingly need to “bake ethical responsibility into the infrastructure”—embedding guardrails directly into workflows and systems from the start rather than applying them via prompts at the front door and after deployment.
Human oversight needs evidence
Boinodiris warned that simply placing a human in the loop isn’t enough if that person doesn’t understand the AI system. For high-risk uses, humans need observability, data provenance, evidence and the ability to ask why a system produced an output. Without that, she said, human oversight can become “liability laundering.”
The broader sovereignty question is whether companies can prove control in real time: who controls the system, how it operates and whether controls hold across data, AI, identity and operations. This is beyond where workloads run. It’s about whether evidence is continuous, whether accountability is clear and whether governance is built into the system rather than reconstructed after an incident.
The companies that engineer control into speed will pull ahead
For consumer goods companies, the payoff isn’t slower, safer AI. It’s faster AI that enterprises can trust.
Governance, done well, allows leaders to deploy more agents, see costs clearly, intervene when systems drift and scale use cases without reinventing the controls every time. It also helps ensure that the underlying data—the fuel AI depends on—is clean, interoperable and fit for purpose.
As AI becomes embedded across the value chain, the winners will be the ones that engineer control into speed—and make sure their AI, like any other powerful system, has earned its license to operate.
Business users want new applications now. Market and regulatory pressures require faster application updates and delivery into production. Your IBM i developers may be approaching retirement, and you see no sure way to fill their positions with experienced developers. In addition, you may be caught between maintaining your existing applications and the uncertainty of moving to something new.
IT managers hoping to find new IBM i talent are discovering that the pool of experienced RPG programmers and operators or administrators with intimate knowledge of the operating system and the applications that run on it is small. This begs the question: How will you manage the platform that supports such a big part of your business? This guide offers strategies and software suggestions to help you plan IT staffing and resources and smooth the transition after your AS/400 talent retires. Read on to learn:
LATEST COMMENTS
MC Press Online