From facilitator to orchestrator: Why the consumer goods CIO is becoming the architect of AI-driven growth
How CPG CIOs can turn AI into trusted, governed business execution
Consumer goods companies have moved past the AI debate. Now comes the harder part: leadership. As enterprise-scale deployment becomes the new baseline, the CIO’s role is undergoing a fundamental rewrite.
According to Gartner’s Top Trends for Consumer Goods CIOs in 2026 report, consumer goods CIOs are expected to focus on modernizing technology, processes, and governance to support trusted AI outcomes.
CIOs must ensure AI delivers trusted business outcomes rather than simply faster answers.
To explore what that transformation means in practice, we spoke with Alessio Bonfietti, Chief Science Officer at XTEL, about the future of AI in CPG, the limits of chatbot-centric thinking, and why the next wave of value creation will come from orchestrated, domain-specific intelligence.
Key Takeaways:
The CIO becomes the orchestrator
AI value depends on governed data, workflows, and decision-making across the enterprise—not just new tools.
Stop building “chat-first” AI
The goal is reliable execution: AI that completes business tasks with humans in control, not just answers questions.
Architecture beats model size
Combine domain context with deterministic systems for calculations so outputs stay correct under real commercial constraints.
Q1: Gartner’s report suggests the CIO role is evolving from a technology facilitator toward business orchestrator. Does that match what you’re seeing?
Alessio Bonfietti: Absolutely. A decade ago, if your systems stayed online, your security patches were up to date, and the ERP didn’t crash, you had a great year. Those responsibilities didn’t go away—but today, they’re just table stakes.
Now, AI touches nearly every part of the enterprise: commercial planning, revenue growth, supply chains, finance, and customer relationships. As a result, the CIO role is evolving beyond technology operations toward orchestrating trusted data, governance, and intelligent systems across the business. Not every organization is there yet, but many are moving in that direction.
It’s less about deploying new software and more about orchestrating how the entire business makes decisions.
Q2: Many organizations still treat AI like a glorified search bar. Is that limiting progress?
Alessio Bonfietti: In many cases, yes. I think we’ve become a little obsessed with interface novelty instead of business outcomes.
The fatigue isn’t about the technology. Instead, it’s about the reliability gap. Users are tired of tools that give different answers to the same question—or that ask them to verify every output. That’s not assistance; that’s extra work with extra steps. People are tired of the homework.
Real enterprise value doesn’t come from asking a chatbot better questions. It comes from AI quietly helping work get done. The conversation is shifting from “What can AI tell me?” to “What can AI do for me?”
That’s where agentic AI becomes genuinely valuable. Not because it’s another model, but because it can orchestrate workflows while keeping humans firmly in control. Natural language chatbots must progress from “Here is how to do this” to “Task complete.” That’s the critical shift from AI systems that primarily provide information to those that help complete business tasks.
“Enterprise AI isn’t a trivia competition”
Q3: There is a lot of focus on choosing the right AI model. But is the model itself really the most important decision for enterprise AI?
Alessio Bonfietti: Not at all—and this is one of the most important distinctions we need to make clearly. The model matters, but the architecture around it matters more.
General-purpose large language models are incredibly smart, but enterprise AI in consumer goods isn’t a trivia competition. The challenge isn’t just generating an answer. It’s generating the correct answer under strict commercial constraints, for a specific customer, product, region, and planning cycle.
For example, if an AI suggests a brilliant 20% discount on a laundry detergent line during a holiday weekend, but fails to account for your retailer margin agreements, localized supply bottlenecks, or pricing rules, that fluent answer is completely useless.
In CPG, the business questions that matter are rarely open-ended. They’re highly specific: How does promo uplift behave for this retailer in this region during this season? How do cannibalization and halo effects interact at category level? How do trade terms, net prices, and customer P&L connect? What does the right baseline volume look like before we model incremental?
A general-purpose model may understand what a promotion is, but enterprise performance depends on much more than what the model knows. It depends on whether the AI has access to the right commercial context, understands the workflow, can invoke the right systems and calculations, and operates within the right business constraints.
The question, therefore, is not simply whether you use a large or small model. It is whether you have designed a system that combines the appropriate model with domain knowledge, enterprise data, deterministic tools, and governance.
This is where smaller, domain-adapted models can become very compelling. For well-defined tasks, they may offer lower latency, lower cost, and greater control. However, the goal is not to use the smallest model possible—or the largest. It is to use the right model for each part of the workflow.
In a mature enterprise AI architecture, there may not even be one model. Different models can be routed to different tasks, while deterministic systems remain responsible for calculations and governed business logic.
Ultimately, the next wave of commercial AI will not be defined by model size. It will be defined by architecture: models with the right domain context, connected to the right enterprise data and tools, operating within governed workflows. The model is an important component—but it is not the system.
“Stop treating LLMs like calculators”
Q4: If AI shouldn’t do everything, what does a successful enterprise architecture actually look like?
Alessio Bonfietti: One of the biggest mistakes we see is companies trying to use AI to replace every single system they own. The opposite is what actually works.
Consumer goods companies already have forecasting engines, optimization solvers, pricing simulators, trade promotion systems, approval workflows, and financial calculation layers. These deterministic tools produce highly reliable, governed outputs. They do math beautifully. The worst thing you can do is ask a language model to replicate what those systems already do well.
The better architecture starts with understanding what the user wants to accomplish. A domain-specific language model interprets that intent and determines which enterprise systems or calculation engines should perform the work. Those deterministic tools then execute the calculation, producing a governed, auditable result. Finally, the AI presents the outcome in clear business language, explaining both the recommendation and the reasoning behind it.
The AI layer acts as the intelligent interface between the business user and the system of record. The deterministic engine still performs the calculation. The model makes the workflow easier, smarter, and more contextual, without replacing the governed systems that calculate forecasts, optimize plans, or enforce financial constraints.
AI’s job shouldn’t be to replace those tools; it should be to connect them. It should interpret what a user wants to do, call the right calculation tools, explain the math in plain business English, and help the user make a call. That’s a highly trustworthy system. Expecting a raw language model to calculate your trade promotion ROI on its own is a recipe for disaster.
This hybrid approach, by moving critical calculations outside the probabilistic model, dramatically reduces the risk of hallucination in calculation-intensive workflows. The model explains the trend and provides context; the deterministic engine delivers the precise 4.2% lift forecast. The human no longer has to double-check the math—they review the strategy. That’s where real operational value is unlocked.
At XTEL, this is exactly the kind of architecture we’ve built around. Our revenue management suite combines domain-specific commercial intelligence with deterministic optimization workflows, so CPG companies get the explainability of AI with the precision and auditability of governed systems.
Q: As AI becomes more autonomous, what keeps you up at night?
Alessio Bonfietti: Governance. Without question.
Everyone wants faster decisions, but speed without trust creates risk. The companies moving fastest with AI are usually the ones that first invested in data quality, governance, and clear business processes. AI amplifies whatever foundation already exists. If the underlying data and processes are messy, AI will only help you make bad decisions at lightning speed.
For CIOs, the role is to make sure the foundation is rock-solid for AI to scale responsibly. That means modernizing integration layers, automating workflows, embedding cybersecurity frameworks, and—critically—managing organizational change. Modernization reshapes business processes and demands significant talent reskilling.
The implication of Gartner’s 2026 report is clear: without modernized technology stacks, processes, and governance, organizations risk scaling bad decisions just as quickly as good ones. The governance infrastructure is not a constraint on AI adoption. It’s what makes AI adoption trustworthy.
Final advice for CIOs
Q: If you could give every consumer goods CIO one piece of advice, what would it be?
Alessio Bonfietti: Stop treating AI like an IT project. Treat it as a business transformation initiative.
Start with workflows that directly affect growth and profitability, such as revenue growth management, promotion planning, trade spend optimization, and demand forecasting. Build around measurable outcomes: faster planning cycles, reduced escalations, better straight-through processing, improved trade ROI, and higher-quality commercial decisions.
And fix your data, and build guardrails into every single step.
The organizations creating competitive advantage aren’t necessarily the ones using the biggest AI models. They’re the ones embedding intelligent, domain-specific decision-making into everyday commercial processes—and governing it well enough to trust the output.
Closing Thoughts
One of the key themes in Gartner’s research is that AI alone will not transform consumer goods organizations. Success depends on modernizing technology, data, governance, and business processes so AI can deliver accurate, actionable insights at scale.
And the model matters. Generic, large-scale language models are powerful general tools, but CPG and retail workflows require precision, domain knowledge, and commercial logic that general-purpose models cannot reliably provide on their own.
The future belongs to domain-specific AI architectures: systems that combine the appropriate models with enterprise context, deterministic tools, governed data, and the workflows where commercial decisions actually happen.
Competitive advantage will not come from having the largest model, or even from having a proprietary model. It will come from how effectively companies compose intelligence, data, tools, and business logic into trusted decision-making systems.
In Bonfietti’s view, the CIO’s role is no longer to simply enable technology. It is to orchestrate the people, processes, and platforms that allow AI to become a trusted part of how the business operates.
Bonfietti argues that organizations will differentiate themselves not by deploying the most AI, but by embedding trusted intelligence into business processes. The most successful CIOs will be the ones who embed intelligence into the moments where commercial decisions are made—and ensure that intelligence is trustworthy enough to act on.