AI reality check: Sustaining the AI growth dividend: PwC

Dan Priest

Chief AI Officer, PwC US

Email

Sean Viergutz

Chief Clients and Markets Officer, PwC US

Email

Key takeaways:

AI’s growth dividend is already taking shape despite fears of an economic reckoning. Productivity, revenue and earnings are showing gains, even as disruption, including fears of a “SaaSpocalypse,” underscores how AI could challenge traditional business models and force industries to reinvent themselves.

The best defense against AI's worst-case scenarios is building the capability to manage them. Organizations that build technical expertise, governance, independent evaluation, and accountability can manage emerging risks while continuing to innovate.

Capturing AI's growth dividend requires more than investing in technology. Learn what leaders should do now to manage risks, build trust, and turn AI's potential into new opportunities for growth.

The headlines suggest AI is heading toward a reckoning. Warnings of a jobs apocalypse, a SaaSpocalypse, an economic reckoning, and powerful new cyberthreats dominate the conversation. 

But look beyond the headlines, and a different picture emerges. AI investment is helping fuel economic growth, lift corporate revenue and earnings, and accelerate productivity. The growth dividend is already taking shape. The question is whether businesses can sustain it. 

We don’t believe leaders need to choose between taking AI’s risks seriously and moving their businesses forward. In fact, the two are increasingly inseparable. It’s reasonable to have questions about AI’s reliability, the security of your data, and what increasingly autonomous systems mean for your people. Those questions deserve thoughtful answers. And a plan. 

AI is maturing into a general-purpose technology at an unprecedented pace, creating enormous opportunities—and risks—for businesses. 

The blistering pace of this maturation process is forcing management teams to handle the transformation of major enterprises in compressed time frames, moving toward a future of new possibilities before they fully understand its implications. AI can make cyberattacks cheaper and faster. Agents can act outside their instructions. Organizations are building dependencies on systems they may not fully understand.  

At the same time, AI is already creating extraordinary opportunities to reimagine products and services, bring new ideas to market faster, and create entirely new sources of growth. 

Disruption is real. So is growth. 

If AI were broadly destroying economic value, we’d expect to see it in the fundamentals, including output, revenue, earnings, productivity, and employment. Since 2023, those first four have improved, while unemployment has risen modestly before stabilizing. 

Disruption is happening, and its effects can be significant for individual workers and companies. But we need to distinguish those disruptions from the broader economic trajectory. Here’s how four of the loudest fears hold up against the evidence. 

Fear No. 1: A jobs apocalypse 

Reality: US payroll employment has grown by 2.2 million jobs since the end of 2023, but unemployment has also risen. AI’s role in labor market disruption remains difficult to isolate and the new categories of work it may create are hard to predict. The question for leaders is how to use AI to augment human capabilities, redesign work, and create opportunities for their people. 

Fear No. 2: AI spending is outrunning the economy 

Reality: Information-processing equipment and software, the category carrying much of AI capital investment, accounted for approximately 72% of real GDP growth in the first quarter of 2026. Earnings are projected to grow 24% and productivity growth is trending above the previous decade’s averages. Longer-term returns remain uncertain, but the economic contribution is already measurable. 

Fear No. 3: The SaaSpocalypse 

Reality: AI is challenging traditional software business models, but disruption isn’t destruction. Despite sharp valuation declines in the first half of 2026, companies including Salesforce, Workday, and Adobe continued generating revenue and operating income growth with AI-driven offerings becoming important contributors. Software valuations remain volatile, and share prices fell again in late October 2026, but results have held up so far. AI-powered code generation is among the first AI capabilities to be industrialized at scale, threatening established software business models and offering a preview of the disruption other industries may face. 

Fear No. 4: AI investment isn’t paying back 

Reality: PwC’s 2026 Global CEO Survey found that 44% of respondents are already seeing financial benefits from AI, while our AI Agent Survey found 88% of enterprises are increasing AI budgets. Results remain uneven, but there are leaders who have honed their approach to get the outcomes they want. And enough organizations are demonstrating value that the question is shifting from whether AI can deliver returns to how to capture more value. 

Trust is the next growth accelerator 

One of the biggest risks we see isn’t that AI capability will stall. It’s that adoption won’t keep pace. 

Models are becoming more capable, inference costs are falling, and autonomous systems are moving into increasingly consequential workflows. But organizations can’t scale what they don’t understand, can’t control, or can’t defend to employees, customers, and boards. 

Public attitudes underscore the challenge. Surveys consistently find that only about one-third of US respondents trust AI, compared with more than 85% in China. That matters because trust influences whether people embrace AI and whether organizations can embed it into the work that creates value. 

There’s a difference between AI that’s used and AI that’s adopted. Used AI sits on the side of the business as isolated pilots and productivity tools. Adopted AI becomes part of how the business operates, shaping decisions, workflows, and customer experiences. 

Making that transition requires confidence that AI systems will behave reliably, that risks are understood, and that someone is accountable when things go wrong. 

Trust isn’t something organizations can declare. It’s something they have to engineer, demonstrate, and sustain. 

What deserves the worry—and how to manage it 

AI risks are real because of how powerful these systems have become. Agents can execute multistep tasks, interact with enterprise systems, and take actions with limited human intervention. Recent incidents have raised questions about rogue agents operating beyond their intended boundaries, potentially accessing or compromising systems without authorization. As organizations connect these capabilities to critical processes, the consequences of unexpected behavior grow. 

But these risks aren’t reasons to retreat from AI. Existing laws already establish accountability for products and services, including AI. Unauthorized access to another company’s systems, whether carried out by a person or an AI agent, can have serious legal consequences. Managing them requires more than technical safeguards. Organizations need assurance that AI systems are operating responsibly, that the cyber and engineering capabilities to secure and monitor them exist, and that the business strategy and organizational change needed to build confidence and trust in AI will work. Together, these capabilities build the confidence needed to adopt AI and realize its value. 

The important distinction is between risks inherent in the technology and those introduced by how organizations design, integrate, and deploy it. Both require attention, but they also point to practical solutions. 

Four areas deserve particular attention.  

Critical dependencies without visibility into failure: AI systems may fail because of immature design, poorly validated integrations, or unexpected actions. Leaders need to understand which systems are business-critical, what dependencies they introduce (building in consideration of AI maturity, and factors such as task length and complexity), and how failures will be managed. 
Agents acting outside their instructions: Autonomous systems introduce concerns about reliability, security, explainability, and safety. Model providers must strengthen these capabilities while enterprises need independent evaluations, permission boundaries, monitoring, and mechanisms to contain unexpected behavior. 
Security threats at machine speed: AI can accelerate familiar cyberattacks while agentic systems expand the attack surface through connections to enterprise data and applications. Organizations need to secure integrations, limit permissions, test for misuse, and monitor agent activity. They must also prepare for a world of AI-on-AI cybersecurity, where AI-powered defenses are critical to detecting and countering threats operating at machine speed. 
Incidents without clear accountability: Companies need to adjust and confirm accountability models as AI systems become more autonomous. A lack of clarity around ownership, reporting, and incident response can have unintended consequences.

Recent incidents illustrate the challenge of AI agents operating without adequate sandbox security and monitoring, and AI-enabled attacks exploiting vulnerabilities previously considered practically unexploitable. The lesson isn’t that organizations should slow down AI innovation but that they need to strengthen their ability to manage its risks as capabilities advance. 

Preparedness is a capability, not a posture 

The best defense against AI’s worst-case scenarios is building the capability to manage them. AI will continue to advance whether individual organizations move forward or not. Slowing down doesn’t necessarily reduce the risks they face. It can leave them less equipped to understand, anticipate, and respond to those risks. Organizations that invest in the technical expertise, governance, and controls to deploy AI responsibly will be better positioned to manage its risks—and capture its opportunities. Independent evaluation is particularly important. Testing systems against defined performance, safety, and security expectations helps identify weaknesses before deployment. And because models, integrations, and workflows evolve, evaluation and monitoring must be continuous. 

Public policy is also playing its part, and more is on the way. Existing laws already make companies liable for their products and services, and that includes AI. States are enacting AI-specific guardrails, multiple executive orders are in effect, and new federal legislation is under consideration. Of course, waiting for policy to create trust in AI isn’t good business. The White House Accord is voluntary: Companies are choosing to adopt its measures before any law requires them to.  

The economics of AI reinforce the urgency. Enterprise AI spending is now six times its 2023 share of revenue, while inference prices have fallen a hundredfold for fixed capabilities. The Jevons paradox suggests that as a resource becomes cheaper, demand for it can increase rather than decline. As AI becomes more affordable and widely used, organizations that invest in preparedness will be better positioned to manage its risks and capture its opportunities. 

The AI growth dividend 

Building trust is essential to accelerating AI adoption. One way to do that is to refocus on a growth and innovation agenda with safety at its core.   

AI creates capacity—hours returned, expertise made scalable, processes redesigned, and new capabilities available at lower cost. But the growth dividend comes from what organizations do with that capacity. 

Too much of the AI conversation centers on reducing headcount or making existing processes cheaper. The bigger opportunity is to grow the pie by creating new jobs, reimagining products and services, entering new markets, and pursuing opportunities that weren’t possible before. 

As inference costs decline, more applications become economically viable. But capturing those opportunities requires more than deploying tools. Organizations need to redesign work, build capabilities, and invest in the controls that make scaled adoption possible. 

PwC’s benchmarking shows that companies consistently investing heavily in AI are already pulling ahead. They have an EBITDA growth advantage of up to 19 percentage points in some industries, revenue leadership in four of five industries examined, and shareholder returns 21% above the typical company in their sector. 

Investment helped open that gap. We expect preparedness and trust to help determine which organizations sustain and widen it. Companies that build safety, security, and accountability into deployment from the beginning can scale what others are still trying to validate or defend. 

What leaders need to get right now 

We don’t need leaders to be fearless about AI. We need them to be prepared to move forward responsibly. That means guarding against worst-case scenarios without letting them become reasons to forgo AI’s potential. 

New categories of work, along with products and services, will emerge. Discoveries in medicine, materials, and energy could accelerate, creating opportunities for growth and prosperity that we’re only beginning to imagine. The organizations best positioned to capture those opportunities will be those that build the capabilities to innovate safely and responsibly. 

That’s how AI’s growth dividend becomes real. Here are the decisions that matter now. 

Make safety part of your strategy. Emerging standards for AI governance, safety, and trust are rooted in reasonableness. Know which AI systems and agents are operating, what they’re authorized to do, who owns them, and what happens when something goes wrong. Build governance, testing, monitoring, and incident response into deployment from Day One, and fund them. 
Build trust through action. Give employees and customers confidence in how AI is used through transparency and accountability. Ground those actions in a strategy that articulates how people and AI can work together to create value and achieve what wasn’t possible before. 
Test the headlines against your fundamentals. Where is AI creating value for your business, your teams, and your customers—and where isn’t it? 
Find the growth, or account for its absence. Defense is a posture, not a growth strategy. Decide where AI will help you lead. If benefits aren’t showing up in performance, new products, or new ways of working, either the curve hasn’t reached you or your strategy hasn’t positioned you to capture those benefits. 
Fund the transformation, not just the tooling. Tools are the easy part. The dividend is earned through redesigned work, new capabilities, and new growth with safe deployment funded from the start.