Key takeaways

• AI is now table stakes: 87% of organizations are applying GenAI or Agentic AI in IT operations, yet the average expected failure rate for major AI projects initiated over the next 24 months is 43%. The gap between ambition and impact is a defining competitive question for retail and consumer packaged goods (CPG).
• The barriers are not purely technological:Legacy applications, siloed data, cross-functional friction, skills gaps and misaligned leadership expectations are among the structural barriers organizations need to address.
• Three imperatives stand out: Modernize foundations, operationalize Responsible AI and select the right partners. Organizations working with expert partners outperform those going it alone across speed, impact, cost and Physical AI maturity, including more than twice the rate of Physical AI production deployment (54% vs. 26%).

Somewhere right now, a retail CIO is walking into a boardroom with an AI roadmap under one arm. The board expects fast payback. The CIO knows there is more risk behind that expectation than the roadmap may suggest.

HCLTech's enterprise AI market report, The AI Impact Imperatives, 2026, found that respondents expect 43% of major AI projects initiated over the next 24 months to ultimately fail. 

They're presenting the roadmap anyway. And they're right to.

To understand why, remember the last time retail faced a shift of this scale. Retail has seen a version of this story before. It was called ecommerce.

By the turn of the millennium, retailers were racing to establish an online presence. But having a website was not the same as building a viable ecommerce business. The URL was the easy part. What was harder was everything behind it: inventory that told the truth, fulfillment that could pick single units and a profit and loss model that stopped treating online as a side project.

The retailers that ultimately built lasting advantage did more than adopt the channel. They reshaped operations around it.

AI adoption is no longer the differentiator

Our research suggests AI has reached a similar moment. Adoption is no longer the story:

• 87% of organizations are applying GenAI or Agentic AI in IT operations 
• 86% are applying these technologies in software development 
• 79% are using AI in physical environments and production operations 

As the report puts it, ubiquity is the baseline, not the headline. The question now is not simply who has AI, but who can translate widespread adoption into measurable business value.

Kristina Rogers, Chief Growth Officer, Retail, CPG and Luxury at HCLTech, puts the shift in perspective:

“Much like e-commerce transformed retail a generation ago, AI is reshaping how organizations operate, compete and grow. The next generation of leaders will be defined not by AI adoption alone, but by their ability to embed intelligence across the enterprise and turn innovation into measurable business value.”

And this time, the transformation comes with a demanding payback clock.

AI spending is accelerating

Across the organizations surveyed, the mean share of technology spending allocated to AI is poised to increase by roughly 32% from current levels over the next 12 months. The proportion of respondents who estimate that more than 20% of their technology investments will support AI projects and initiatives rises from 32% over the past 12 months to 48% over the next 12 months. 

That's not experimentation money. It represents a significant shift in how enterprises are funding their technology priorities.

The spending follows a hard operational reality. Agentic AI, meaning systems that can reason, plan and perform actions, is already executing IT operations work according to 78% of IT leaders surveyed, while 59% of operations leaders say it is helping with production operations. 

The research found that 15% of respondents say 100% of competitors in their industry already use autonomous AI systems for mission-critical work. When those expecting this within the next 12 months are included, the figure rises to 77%. 

In 2025, Omdia also forecast that the Agentic AI software market would grow from $1.5 billion in 2025 to $41.8 billion by 2030. 

Vijay Guntur, Chief Technology Officer and Head of Ecosystems at HCLTech, describes the moment plainly:

“AI has moved from being a technology initiative to becoming an enterprise operating reality. What leaders are grappling with now is not whether AI can deliver value, but how organizations adapt their structures, decision rights and risk tolerance to keep pace with it.”

For retail and consumer packaged goods, the bar for all that new spending is unambiguous. Corrado Azzarita, Global CIO of The Kraft Heinz Company, said:

“At the end of the day, only three metrics matter in business: revenue, cost and risk. If a provider cannot clearly articulate how a solution impacts at least one of those, then everything else is just noise.” 

Leaders need a fuller understanding of AI-related risk

We asked the leaders responsible for advancing AI, including chief AI officers, chief information officers, chief technology officers, chief data officers and their teams, how well their CEOs and boards understand what AI leadership requires.

• 87% said their CEO and board have significant gaps in understanding that AI carries a high level of investment risk and that not every initiative will bear fruit 
• 85% said their CEO and board need material improvement in understanding that leading on AI may require some medium-term margin pressure due to capital outlays 
• 83% said their CEO and board do not adequately understand that underinvesting in AI may create existential risk for the organization 

These findings point to a broader leadership understanding gap. Leaders need a clearer view of what successful AI adoption requires, including investment risk, realistic return expectations and the organizational changes needed to scale AI.

The median expected payback period for major AI initiatives is roughly 18 months. 

For retail and CPG organizations managing tight margins, seasonal peaks, promotional cycles and rapidly changing consumer demand, that creates a difficult equation: an aggressive return timeline set by leaders who haven't priced the risk, alongside an average expected failure rate of 43% for major AI projects initiated over the next 24 months.

Jill Kouri, Global Chief Marketing Officer at HCLTech, identifies what can sit behind those failures:

“Every AI initiative is, at its core, a change management initiative. The technology is rarely what fails. What fails is the human architecture around it - the behaviors, the trust, the willingness to work differently.”

To understand how organizations can change outcomes, leaders need to look beyond individual projects to the systems and the people standing between the lab and the shelf.

Legacy foundations are holding AI back

This is where promising AI initiatives can stall. Not necessarily in the model itself, but in the aging systems between them.

When we dug into what limits AI adoption, the top technical culprits weren't models. They were: 

• Security vulnerabilities unique to AI systems (34%)
• Integration challenges between multiple AI systems (30%)
• Data quality, quantity and accessibility (29%)
• Technical debt and difficulty modernizing legacy applications and data sources (28%)

A shortage of internal AI expertise ranked lower, at 22%.

Across the functions surveyed, respondents estimate that roughly half of applications in production are legacy. The research found that 96% say those applications are creating challenges such as high management costs, performance limitations, integration issues, security vulnerabilities and slow development cycles.

Only 21% say their data estate is modernized and operational. 

For retail and CPG, the implications are easy to recognize: a merchandising platform installed decades ago, a point-of-sale system that cannot exchange data in real time, an enterprise resource planning platform disconnected from newer applications or trade-promotion tools operating on fragmented information.

The same issue can emerge in demand forecasting, retail media, inventory planning and supply chain operations. AI may be capable of acting in real time, but it cannot make good decisions when it lacks reliable access to data. On the CPG side, demand sensing may lack access to retailer data, while retail media can be forced to work from stale signals.

Rebuilding that foundation carries a cost. The research models the labor cost of a typical application modernization at approximately $807,000, based on a median of 8.5 full-time equivalent developers and approximately 39.5 weeks of work. The actual cost will vary depending on the application, its dependencies and the complexity of the environment. 

But standing still has a cost too: 52% of organizations say they are already at a competitive disadvantage because they cannot modernize applications fast enough. 

Pawan Vadapalli, Corporate Vice President and Global Head, Digital Business Services at HCLTech, frames the stakes and the opportunity:

“When half your application estate was built before the modern AI demands existed, you're not just carrying technical debt, you're operating under an AI performance ceiling. The shift comes when you realize the same AI that requires modern infrastructure is also the fastest way to build it. What once took four years can now take 12 months or less.”

That connection between modernization and AI becomes important because organizations are trying to pursue both at once.

Some of the biggest barriers to AI adoption are organizational

Business and IT are pulling in different directions

• 62% of respondents agree that business leaders are frustrated by what they perceive as slow delivery of high-profile AI projects by IT 
• 68% say IT leaders are concerned about business teams advancing AI projects without proper oversight and governance 

Both concerns are legitimate. Together, they can create a difficult cycle: business teams push ahead because they see governance as slowing progress, while IT becomes more cautious because uncontrolled deployment increases risk.

Responsible AI concerns are already influencing delivery. 76% of respondents report that their organization has frequently or sometimes delayed AI deployments due to Responsible AI concerns. 

But the research also shows why the answer is not to reduce governance.

And the gap runs deeper than the org chart. Cross-functional coordination (40%) and managing organizational change and resistance (30%) both rank ahead of securing adequate funding (26%). The report also identifies skills gaps as a barrier, reinforcing the need to address people and organizational change alongside the technology.

This is the low point of the story. It's also exactly where the data turns.

Organizations with very mature Responsible AI practices report substantially fewer difficulties balancing AI innovation with ethical considerations than organizations with immature approaches: 24% vs. 41%. 

Responsible AI, done well, does not simply mitigate risk. It can help create the confidence required to deploy AI more widely.

For consumer-facing brands, that is important. Personalization, automated decision-making and intelligent customer experiences all depend on data and trust. Responsible AI can become a competitive differentiator, not simply a compliance function.

Three imperatives to improve the odds

The 43% failure expectation should not be read as a verdict on AI itself. The research points instead to structural challenges around foundations, business alignment, governance and organizational readiness.

Address those conditions and organizations are better positioned to turn AI adoption into measurable impact.

Our research points to three imperatives.

1. The right foundation

   Organizations are also turning to AI for the modernization challenge itself.

   The research found that 75% of respondents say they will rely on AI tools to unlock the productivity and efficiency needed to achieve their application modernization goals. Four in five respondents also agree that both legacy and modernized applications will require major rewrites to meet AI demands.

   That means leaders do not have to treat modernization and AI adoption as separate priorities. AI can help accelerate the development of the foundation on which it ultimately depends.

   For retail and CPG, the payoff could appear in areas such as demand forecasting that sees a more complete signal, inventory decisions that respond faster and edge intelligence capable of acting on operational data closer to the shelf, store or factory.

2. The right governance

   Responsible AI must move from policy into day-to-day operations.

   Organizations with very mature Responsible AI practices report fewer conflicts between innovation and ethical considerations, as well as fewer challenges around accountability, transparency and Responsible AI policies. 

   For retail and CPG organizations handling consumer, employee, product and commercial data, this can make trust an enabler of AI adoption rather than a barrier to it.

3. The right partners

   This is one of the clearest patterns in the research.

   Across speed, impact, cost and Physical AI maturity, organizations working with consultative partners to advance AI outperform those going it alone. Among organizations already using external AI expertise, 90% say partners accelerate ramp-up and time to value, 89% say partners increase the business impact of initiatives reaching production and 86% say partners help manage AI-related costs. 

   The difference also extends into Physical AI. Organizations working with external partners are more than twice as likely to have Physical AI deployed in production, at 54% vs. 26%. 

   These capabilities are particularly relevant to retail and CPG. Across the overall research sample, respondents identify logistics and supply chain management (63%) and manufacturing (55%) among the operational domains that could benefit from Physical AI. Among organizations that have already deployed it, 62% report improved production-line uptime. 

   Hari Sadarahalli, Corporate Vice President and Global Head of Engineering and R&D Services at HCLTech, explains the difference:

   “Organizations achieving the greatest impact are those that treat AI as an engineering discipline, not a technology experiment. That rigor is what transforms proof of concepts into scalable, enterprise-grade systems.”

   Through HCLTech AI Labs, we have hosted hundreds of clients across industries and at different stages of their AI journeys, helping them navigate these challenges.

What the board will ask next

The CIO does not need to open with the roadmap. They can open with the odds: 43%. Then comes the question that reframes the discussion: Are we going to treat AI as another technology deployment or rebuild the organization around what it makes possible, as the strongest retailers did with ecommerce?

The roadmap is how those conditions begin to change.

And the full report provides benchmarks that can help shape that conversation. Among them:

• Organizations that are most successful and partner with third-party AI experts allocate 31.2% of their technology budgets to AI, compared with 24% among those going it alone.
• Four in five respondents say both legacy and modernized applications will need major overhauls to meet AI demands. 
• Physical AI findings show a strong preference for company-specific, fine-tuned models and purpose-built tools over free, general-purpose AI models. 
• When organizations select AI partners, they look beyond AI expertise to capabilities in adjacent areas, including cloud, application modernization, security and compliance. 

The 43% represents an average expectation across respondents, rather than a prediction for every organization, and signals the scale of execution risk.

For retail and CPG leaders, the opportunity is to understand what sits behind that risk and build the foundations, governance and partnerships needed to turn widespread AI adoption into sustainable business impact.