51% of AI projects miss targets, yet 85% of firms raise budgets

Just 49% of AI initiatives funded in the past 24 months delivered the business value companies expected, according to new Genpact research — yet 85% of surveyed enterprises plan to increase AI spending anyway. The study, published November 13, 2026, surveyed more than 2,000 senior executives across 17 industries and 14 business functions, at companies with annual revenues from $1 billion to more than $50 billion, supplemented by in-depth interviews with enterprise leaders.

The headline number is stark: 51% of AI initiatives failed to deliver expected value. The gap between that failure rate and continued budget growth frames the central tension in enterprise AI right now. Boards and senior management keep funding deployments, Genpact found, while still working through unresolved questions about governance, integration and measurement.

How widespread is AI adoption actually?

Adoption is broad but shallow. Genpact found that 79% of respondents have AI agents (autonomous software systems that perform tasks with limited human input) in production. Only 20%, however, said they use those agents broadly and consistently across functions. A mere 9% reported completing significant operating model transformation — the restructuring of workflows, management and accountability that turns tools into financial results.

AI now leads or fully runs 40% of decisions inside surveyed companies, more than double the level a year earlier, according to Genpact. That figure carries a caveat the report itself supplies: humans still make or audit decisions in at least 87% of cases at every stage. Autonomy on paper remains heavily supervised in practice.

Takeaway: most enterprises have deployed AI somewhere, but few have changed how the business actually operates around it.

Why does deployed AI fail to produce value?

Only 15% of respondents said they had fully linked AI investment to profit-and-loss outcomes. Genpact argues the primary barrier is operational, not technical. Many companies adopted AI tools faster than they redesigned the workflows, management structures and accountability systems around them.

The survey data supports that diagnosis. Deployments are widespread, yet organisational reach is limited and transformation is rare. Systems go live; balance sheets stay untouched. Genpact frames this as a distinction between putting AI into use and changing operating models so deployments affect financial performance.

Sanjeev Vohra, Chief Technology and Innovation Officer at Genpact, said the challenge has moved beyond deployment. "Enterprises have become effective at deploying AI agents. Running them is the hard part," he said. "Agentic Operations connects AI with context-rich process intelligence, human judgment, and governance to capture the autonomy dividend."

Note what that quote implies: the missing ingredient, in Genpact's telling, is an operating discipline the firm happens to sell. That does not invalidate the survey findings, but readers should weigh the messenger's commercial position. Takeaway: the bottleneck is process redesign and ownership, not model capability.

What separates the companies that succeed?

Genpact identified a subset it calls AI Leaders — 17% of enterprises in its AI Maturity Index. These companies behave differently in specific, measurable ways:

• They are more likely to overhaul work across the entire business, not just individual tasks.
• They define new roles for employees working alongside AI systems.
• They strengthen data foundations before scaling deployments.
• They tie AI gains directly to named owners and P&L lines.
• They govern AI agents across the full lifecycle, not only at the point of deployment — including overseeing how agents act across systems, not merely how they retrieve information.

That last point deserves attention. Most governance regimes check AI systems when they launch. Lifecycle governance means monitoring what autonomous agents do inside other systems over time, a materially harder problem. The Leaders group is small — under one in five enterprises — which suggests these practices remain the exception rather than the rule. Takeaway: the winners differ in management practice, not in access to technology.

Is AI cutting jobs?

The survey found no immediate workforce reduction from adoption. Among respondents with AI agents in production, 92% said they had either maintained or increased headcount. That figure cuts against the common assumption that agentic AI (systems that act autonomously on tasks) translates directly into layoffs.

One survey wave is not a trend, and "maintained or increased" leaves room for attrition-based shrinking that executives would not label as cuts. But the data at minimum shows enterprises are not yet treating AI agents as a substitution for staff at scale. Takeaway: so far, deployment and headcount are growing together, not trading off.

Why does spending keep rising despite failures?

The core paradox in Genpact's data: 51% of initiatives miss expectations, and 85% of companies still plan to spend more on AI. The report's interpretation is that executives continue to see AI as strategically important even when current projects fall short — a bet that early underperformance reflects immaturity rather than dead ends.

That reasoning has a reasonable basis and an obvious risk. The reasonable basis: the 17% AI Leaders cohort demonstrates that value capture is possible with different operating practices. The risk: if the operational transformation the report prescribes — workflow redesign, lifecycle governance, P&L ownership — does not materialise, rising budgets buy more deployment without more return, and the 49% success rate stagnates. The survey measures intent to spend, not willingness to tolerate indefinite shortfalls.

The stakes extend beyond individual firms. Enterprise AI spending is a principal driver of the current infrastructure buildout across the technology sector; a persistent gap between deployment and financial returns is the kind of pattern that eventually forces corrections in capital allocation. Genpact's data says that correction has not arrived yet.

Takeaway: budgets are rising on faith that operations will catch up with deployment — a faith the data tests but does not yet confirm.

What happens next?

The report positions the coming phase of enterprise AI as an operations problem: connecting agents to process context, human judgment and governance, in Vohra's phrasing. The measurable markers to watch are the ones Genpact's own index tracks — whether the 15% of firms linking AI to P&L outcomes grows, whether the 9% that completed operating model transformation becomes a larger share, and whether the 92% headcount figure holds as agents take on more decision autonomy.

For now, the evidence shows enterprises deploying faster than they reorganise. Until that reverses, expect the value gap to persist even as the spending line keeps climbing.