Via Emerce we came across two major AI studies this week: Deloitte's State of AI in the Enterprise and PwC's Global CEO Survey. The numbers seem alarming: more than half see no ROI yet, 84% haven't redesigned work around AI.
But the reports tell a more nuanced story. We dug in.
The headlines vs. reality
Reading the reports superficially, you mainly see what isn't working yet. "More than half see no ROI from AI!" "84% haven't redesigned work around AI!" Sounds dramatic. But flip those numbers around.
What is happening:
1 in 8 organizations already sees both higher revenue and lower costs from AI. These are the frontrunners doing it right.
of CEOs report revenue growth from AI in the past year.
of organizations have moved at least 40% of their AI experiments to production. And 54% expect to reach that level within 3-6 months.
Access to AI tools has increased by 50% in one year: from 40% to 60% of employees.
That 12% of frontrunners is interesting. What are they doing differently? According to PwC, they have stronger AI foundations: a technology environment that enables AI integration, a clear roadmap, formalized responsible AI policies, and a culture that supports adoption.
The proof-of-concept problem (and the solution)
Deloitte calls it "the proof-of-concept trap." A pilot runs with a small team, cleaned data, in an isolated environment. Production requires infrastructure, integration, security reviews, compliance checks, monitoring, maintenance. Two completely different worlds.
The key? Start with the end in mind. Organizations that treat pilots as stepping stones to production, not as standalone experiments, see much faster results.
"If there's no coherent AI strategy, you get pilot fatigue. You chase the next shiny object, under pressure to do something with AI, without a real plan. When something succeeds, nobody knows how to scale. Without a roadmap, a hundred pilots lead to poor results."
AI leader in the healthcare sector
The access paradox
There's a striking tension in the data. Access to AI tools has increased significantly (60% of employees). But of those employees with access, less than 60% actually use it in their daily work. That percentage has barely changed compared to last year.
Buying more licenses isn't the solution. The challenge is activation, not access.
This also explains why some CEOs think they're doing well ("we've rolled out Copilot to everyone!") while the actual impact remains limited.
Successful implementations start with empowered employees who experiment, share early wins, and become internal champions. Top-down mandates alone rarely work.
Redesigning work: the big missed opportunity
84% of organizations haven't redesigned work around AI capabilities. That sounds like a problem, but it's mainly an opportunity for early movers.
AI demands fundamentally different operational models. A credit analyst who always relied on experience and judgment now needs to collaborate with an AI system that makes recommendations. When do you override the AI? How do you explain decisions to customers? What happens to expertise and career paths?
Most organizations (53%) focus on education: making employees AI-literate. Far fewer actually redesign roles, workflows, and career paths. That's the next step where value lies.
Agentic AI: faster than governance
Nearly three-quarters (74%) of organizations plan to implement agentic AI within two years. But only 21% currently have a mature governance model for autonomous agents.
That's a gap that deserves attention. Unlike traditional AI systems that make recommendations for humans to act on, agents take action themselves: making purchases, sending communications, modifying systems.
Organizations making the most progress approach it deliberately: starting with lower-risk use cases, building governance capabilities, and scaling purposefully. Cross-functional governance that brings together IT, legal, compliance, and business.
What frontrunners do differently
Both reports point to the same success factors:
Enterprise scale, not standalone projects
Isolated, tactical AI projects rarely deliver measurable value. Returns come from enterprise-wide deployment consistent with business strategy.
Strong foundations first
Technology environment that enables AI integration. Clear roadmap. Formalized responsible AI and risk processes. Culture that supports adoption.
Apply AI to products and services
44% of frontrunners apply AI to their own products, services, and customer experience. Compared to just 17% of the rest.
Actively exploring new sectors
42% of companies have started competing in new sectors over the past five years. Companies generating more revenue from new sectors have higher margins and CEOs with more confidence in growth prospects.
Where Deloitte and PwC agree, and where they don't
Both reports were conducted independently, with different methodologies and audiences. Yet they reach strikingly similar conclusions on many points.
Where they agree:
The majority sees no financial returns from AI yet, but a vanguard of about 12-30% does
Access to AI tools isn't the problem; activation is
Governance lags behind the speed of adoption
Strong foundations (infrastructure, data, culture) make the difference between frontrunners and laggards
Companies that take a wait-and-see approach demonstrably underperform those that push forward
Where they differ in emphasis:
Deloitte dives deeper into the operational side: the proof-of-concept trap, work redesign, and emerging technology like agentic and physical AI. PwC takes a broader view of CEO priorities: innovation capacity, M&A in new sectors, and geopolitical uncertainty.
Deloitte sees "insufficient skills" as the biggest barrier to AI integration. PwC sees the broader picture of transformation speed as the core issue, the question that concerns CEOs most is: "are we transforming fast enough?"
What's interesting is that both perspectives support the same conclusion: the problem isn't the technology, but the organization around it.
The bottom line
The data shows a clear pattern. Organizations that go furthest and fastest in restructuring their business and operating models outperform their more cautious competitors.
And that last point may be the most important: companies that say geopolitical uncertainty makes them less inclined to make large investments grow slower (2 percentage points) and have lower profit margins (3 percentage points) than their peers.
Waiting is also a choice. But one with costs.
Deloitte State of AI in the Enterprise (January 2026)
PwC 29th Global CEO Survey (January 2026)
From insight to action?
Curious where your organization stands and what the next step is? We'd love to think along.