A new global study from HCLTech has put a number on something most business leaders already feel in their gut: AI is everywhere, but meaningful results are rare.

The “AI Impact Imperatives 2026” report, conducted with research firm Raconteur across 500 enterprise decision-makers, found that 90% of organisations say GenAI and Agentic AI are transforming their workflows. Nine out of ten. But when asked whether that transformation is actually moving the revenue needle, the picture collapses: only 18% say AI is delivering significant revenue impact.

That gap, between operational disruption and business outcome, is the defining challenge of enterprise AI right now.

What the Numbers Actually Say

The study is careful to distinguish between different kinds of AI impact:

• 91% report improved data access
• 90% report productivity gains
• 18% report significant revenue impact

So AI is making people faster, and it’s giving teams access to more data. Those are real wins. But faster processes and better dashboards are not the same thing as growth. The research confirms what sharp operators have suspected: activity-level AI improvements are common, outcome-level AI wins are rare.

This pattern shows up across multiple research streams. Deloitte’s 2026 State of AI in the Enterprise found that 66% of organisations report efficiency gains, but only 20% have actually increased revenue as a result. Only 29% of organisations see significant ROI from generative AI. For AI agents specifically, that number drops to 23%.

Put bluntly: the majority of enterprise AI spend is producing operational noise, not business signal.

Why the Gap Exists

The HCLTech report identifies what separates AI Leaders from the rest. The companies seeing revenue impact are not just adopting AI, they are “systematically converting it into growth, innovation and customer experience advantage.” That word, systematically, is doing a lot of work.

Most AI implementations fail at the last mile. Businesses buy access to powerful models, integrate them into existing tools, and watch utilisation metrics climb. But utilisation is not value. The gap appears because:

AI without process redesign just automates bad habits. If your sales process has leaks, an AI assistant will just help your team work faster at the wrong things. Productivity goes up, pipeline outcomes do not change.

AI without data foundations cannot do real work. Eighty-one percent of enterprise AI projects hit data quality or accessibility problems early. You cannot build a knowledge-discovery agent on top of siloed, inconsistent data. The model is not the bottleneck.

Adoption metrics get confused with impact metrics. Leaders who are rewarded for “AI usage” have every incentive to report that AI is transforming workflows. Fewer have clear accountability for the specific revenue or cost outcomes.

The Businesses That Are Getting It Right

The HCLTech report identifies a cohort of “AI Leaders” that are pulling away from the pack. What they have in common is not a bigger budget or a more sophisticated tech stack. It is a different approach: they identify specific, high-value workflows, connect AI tightly to those workflows, measure actual outcomes, and iterate fast.

That sounds obvious. But it requires resisting the pull toward broad AI rollouts and platform-level thinking. It means picking the two or three business problems where AI can drive measurable commercial impact, solving those completely, and then expanding.

It also requires a different conversation between AI vendors, implementation teams, and business leaders. The 18% who are seeing revenue impact did not get there by deploying AI across the whole organisation and waiting to see what happened. They got there by treating AI as a business change programme, not a technology programme.

What This Means for Business

If you are in the 90% using AI but not in the 18% seeing revenue impact, the path forward is not more AI tools. It is better targeting of where AI connects to value.

Concretely, that means:

Start with the outcome, not the technology. Define the revenue or cost target first. Then work backwards to which AI capability, in which workflow, with which data, could move that number. Most teams do this in reverse and then wonder why the business case evaporates six months in.

Invest in data foundations. The models are cheap and getting cheaper. The data infrastructure that makes those models useful on your specific business problems is still the hard part. If your data is a mess, AI will be a faster mess.

Measure at the outcome level. Productivity gains are a proxy, not an endpoint. If your AI investments are generating productivity reports but not appearing in your P&L, you are measuring the wrong thing.

Close the skills gap before the model gap. IDC projects that over 90% of global enterprises will face critical AI skills shortages in 2026. The bottleneck is not access to frontier models. It is the ability to understand, direct, and govern what those models are doing in the context of real business operations.

The 18% who are getting this right share a common trait: they have people who understand both AI and the business problems it is supposed to solve. That combination is still rare, and it is the actual constraint.

---

Enterprise DNA exists to close exactly this gap. Through Omni Ops and Omni Advisory, we help businesses design and deploy AI around specific commercial outcomes, not broad adoption goals. Through EDNA Learn, we build the data and AI literacy inside your team that makes those deployments stick. If your AI spend is generating activity but not results, start with a conversation.