90% say AI changed workflows. 18% see revenue.

Here is the most useful number published this week, and it isn't a benchmark. In a survey of 500 enterprise decision-makers, 90% said GenAI and agentic AI are transforming their workflows. Only 18% said AI is delivering significant revenue impact. That's the whole state of enterprise AI in two figures, and the gap between them is where most software budgets are currently disappearing.

What actually happened

HCLTech, working with Raconteur, published The Blueprint for AI Leadership on July 21. Across 500 enterprise decision-makers (press release):

• 90% report GenAI and agentic AI transforming workflows
• 91% report improved data access
• 90% report productivity gains
• 18% report significant revenue impact

The report splits respondents into "AI Leaders" and "AI Followers." Leaders are 4x more likely to have scaled agentic and autonomous AI. The starkest divide is boring and organizational: 93% of Leaders run structured upskilling programs, versus 20% of Followers. HCLTech's Pawan Vadapalli frames the winners as rethinking how the business works rather than bolting AI onto it.

It's a vendor-published survey — a services firm has an obvious interest in an adoption gap it can sell against. Treat the 18% as directional. It still tracks with what we see in the field.

Why the enterprise AI revenue gap matters for your business

Productivity gains that don't show up in revenue usually mean one of two things: the time saved got reabsorbed into other work, or nobody ever measured the baseline, so "productivity gain" is a feeling.

The 18% is not a model problem. Nobody in that survey is failing because Gemini 3.6 Flash needed better benchmarks. They're failing because "we deployed a copilot" was the finish line instead of the starting gun.

What we do before writing a line of automation: name the process, count how long it takes today and how often it runs, and write down the number the automation has to move. Invoice chasing, quote turnaround, order exceptions, intake. If you can't state the current number, you're not ready to automate it — you're ready to measure it. That's not a delay, that's the part that produces the 18%.

The upskilling stat is the other half. An agent that only one person knows how to drive is a single point of failure with a subscription attached. Automations that pay for themselves are the ones your staff can run, inspect, and correct without calling anyone.

Sources: HCLTech, "The Blueprint for AI Leadership" (July 21, 2026), Deccan Herald.