From AI hype to enterprise execution: Insights from Consumer Next

Insights

• AI has moved beyond experimentation into a decisive phase of scaled enterprise execution.
• The biggest challenge isn’t technology – it’s closing the deployment gap through change, skills, and data readiness.
• Real value comes from deep, focused use cases tightly embedded in core business processes.
• Agentic AI enables transformation but only with disciplined design, governance, and architecture.
• Winning organizations treat AI as a strategic capability, not a collection of tools.

Artificial intelligence (AI) is no longer a distant curiosity or an experimental playground: enterprises are rapidly shifting to full-scale execution, moving at a pace unlike any technology wave before. This creates opportunities, but also brings a new set of complex high-stakes challenges.

The Infosys ConsumerNext event, held in London in February with a follow-up event in Düsseldorf in March, brought together industry leaders, technology partners, academics, and practitioners to explore how organizations can accelerate AI transformation at speed and scale, and turn AI potential into tangible business value in consumer goods, retail, and logistics.

The discussion evolved from strategic framing through technology enablement into real-world implementation challenges, culminating in live enterprise use cases.

AI at a tipping point

Ambeshwar Nath, executive vice president at Infosys, highlighted a striking comparison in his introduction: while the internet took a decade to reach a billion users and mobile took five years, AI has achieved similar scale in less than a couple of years. This speed, he emphasized, is unprecedented and places organizations at a critical inflection point.

However, technology is moving faster than adoption and has created what several speakers referred to as a deployment gap — the widening distance between what technology can do and what enterprises are able to adopt. The challenge isn’t as much about access to models or compute as it is about change management, skills, data readiness, and execution discipline within organizations.

From experimentation to differentiation

Drawing on client surveys and recent engagements, the opening keynote noted that 2025 was largely a year of AI experimentation, dominated by pilots and proofs of concept. By contrast, 2026 marks a shift toward differentiation, where chief executive officers and other C-suite leaders are demanding measurable business outcomes rather than fragmented experimentation.

The message was that organizations must focus deeply on a small number of high-value use cases, scale them end to end, and embed them into core operations. AI initiatives that remain horizontal, scattered, or disconnected from business value are unlikely to succeed.

Across consumer goods, retail, and logistics, speakers highlighted tangible value already being realized:

• Revenue growth management powered by AI-driven pricing and promotion optimization in consumer packaged goods (CPG).
• Gross profit improvements of between 3% and 5% in consumer goods through advanced analytics.
• Hyperpersonalized marketing at lower cost and faster speed.
• Computer vision, or using AI systems to interpret images or video to improve how retail stores operate and serve customers, for store compliance and operational excellence in retail.
• Loyalty platforms increasingly driven by AI in retail.
• Logistics optimization, delivering reductions of between 15% and 30% in operational costs.

These examples reinforced a consistent theme: that AI works when it is coupled to business processes and outcomes.

The engine behind enterprise AI

A dedicated session explored NVIDIA’s role in the AI ecosystem, reframing the company not simply as a graphics processing unit (GPU) manufacturer, but as a provider of full-stack accelerated computing platforms.

Sanjeev Arya from NVIDIA highlighted how while most of the world sees NVIDIA mainly as a chip and GPU company, a large part of its value lies in its software and libraries. The AI ecosystem is described as a five-layer stack, starting from power and data centers up through system software, models, and applications. These layers translate traditional workloads so they can run efficiently on GPUs, forming NVIDIA’s accelerated stack.

A key insight was how AI value is realized in applications used by businesses and consumers, and not only in the models. Enterprises are now moving from experimentation to architectural discussions about AI adoption, especially around data readiness, and how NVIDIA’s role is to help enterprises scale, adopt, and operationalize AI, with a focus on real industry use cases like retail.

Why AI is different

The event’s intellectual centerpiece was the keynote by Professor James Fergusson, director of the Infosys Cambridge AI Center. His session provided a rigorous yet accessible explanation of why AI represents a fundamentally different technological revolution.

At the heart of his argument was exponential growth in compute and data. While previous industrial revolutions were driven by finite resources, AI is powered by resources that continue to improve rapidly. This creates repeated phase shifts, where capabilities move quickly from impossible to trivial.

Professor Fergusson introduced the concept of the “bitter lesson”: in the long run, algorithms that exploit compute always outperform handcrafted solutions. This explains why machine learning, and particularly large foundation models, continue to dominate despite their imperfections.

He demystified machine learning, stressing that AI systems are fundamentally function approximators – systems that learn patterns from data and mimic human functions – rather than intelligent beings. Their power lies in scale, not understanding. This leads to predictable strengths, pattern recognition and generation, and equally predictable weaknesses, such as failure outside their training distribution, which refers to situations or data that are different from what the AI was trained on.

From models to systems

A major portion of the keynote focused on agentic AI – systems where multiple AI agents collaborate, plan, execute, and validate work in structured workflows.

Professor Fergusson emphasized that agents are not magical. They require careful engineering, explicit validation layers, and well-designed roles. Without these safeguards, agentic systems become fragile and unreliable.

He shared real research examples where multi-agent systems were used to automate the scientific process end-to-end, from literature review and hypothesis generation to experimentation and paper writing.

As AI develops, increasingly automating specialist technical work, human roles will shift toward managing AI agents, exercising judgment, and collaborating across teams. Professionals will need to combine deep domain expertise with broader, generalist skills rather than relying on narrow specialization. To prepare for this, said Professor Fergusson, education and training should deliberately broaden perspectives, exposing people to diverse disciplines, industry voices, and hands-on experimentation, because single-domain expertise alone will become less valuable as AI advances.

The reality of AI in enterprises

A panel featuring leaders Kuldeep Dudeja from Reckitt, Muhammad Shakir Hussain from Coca-Cola Europacific Partners, and Ben Garside from Currys, moderated by Samad Masood from the Infosys Knowledge Institute, grounded the theory in enterprise reality. The discussion reinforced that AI success is rarely a technology problem but an organizational one.

On whether the AI hype is real, one panelist said that when organizations take a broad, horizontal experimentation approach, it often leads to unclear return on investment (ROI). Instead of spreading AI thinly across many use cases, they adopted a “deep and narrow” strategy. They chose a single process and redesigned it end to end, rather than adding patchy automation or sprinkling AI everywhere. This shift helped them move from experimentation to meaningful impact by focusing on real process transformation rather than tools alone.

Another panelist said AI can be used to generate ROI, but only if you pick the right use cases, and have the right multidisciplinary team and the engineering expertise to build that componentized framework that will sit together and can be expanded and built. He also stated that it’s more of a transformation problem rather that just a technical problem.

It was also discussed that enterprises should avoid chasing “shiny” tools and instead focus on building a scalable AI platform. Commodity capabilities such as standard software-as-a-service (SaaS) integrations can be bought rather than built. Differentiation does not come from the tools themselves, but from how they are assembled and applied. Competitive advantage lies in orchestration and not just ownership of technology. Success depends on bringing together the right tools, skills, data, and organizational mindset. It’s this combination rather than any single software choice that ultimately drives enterprise-scale impact and ROI.

Other candid lessons shared:

• Change management often takes two to three years, far longer than technical development.
• Upskilling is important, but some AI roles also require new talent.
• Saying “no” to low-value AI ideas is essential.
• Board-level sponsorship is a decisive success factor.

The panel also addressed emerging challenges such as shadow IT, data leakage, and rising AI consumption costs. Governance, clear ownership, and enterprise platforms were highlighted as necessary counterbalances to uncontrolled tool adoption.

Platforms, demos, and scaling AI

The latter sessions showcased how AI is being operationalized at scale through platforms and partnerships with companies that have developed industry-ready blueprints for it. Infosys leaders highlighted agent-enabled enterprise platforms designed to orchestrate AI across workflows, embed governance by design, and measure productivity gains.

Live demonstrations included:

• AI-driven store digital twins using computer vision.
• Autonomous document and logistics processing with significant cycle time reduction.
• End-to-end content automation for digital shelves and marketing assets.
• Dramatic cost reductions and compliance improvements through synthetic pipelines.

Across these demonstrations, a consistent message emerged: AI delivers maximum value when embedded into end-to-end systems rather than deployed as isolated tools. This message is underscored by the findings in Infosys’ AI Business Value Radar research report.

From hype to execution

The ConsumerNext event made it clear that AI has crossed the threshold from promise to value. The technology is powerful, accessible, and improving all the time. Success now depends on focus, discipline, partnerships, and people.

Organizations that treat AI as a collection of tools will struggle. Those that treat it as a strategic capability, redesign processes around it, and invest in both technology and change will define the next generation of competitive advantage. The question isn’t whether AI will transform industries, but who will be ready when it does.