An In-Depth Reading and Extended Analysis of PwC's AI Performance Study
Research Basis: PwC benchmarked 1,217 companies across 25 industries globally between October and November 2025, with 76% of respondents reporting annual revenues exceeding US$1 billion. All survey participants held director-level or above positions. Core Finding: The top 20% of companies capture 74% of all AI-driven returns. The most "AI-fit" companies deliver AI-driven revenue and efficiency gains 7.2 times higher than their peers.
The Polarization of AI Value
A More Extreme Version of the Pareto Principle
PwC's research reveals that the top 20% of surveyed companies capture 74% of all AI-driven returns — a distribution even more concentrated than the traditional 80/20 rule of business. This carries two sobering implications:
- Most companies' AI investments are trapped in a "value black hole": Countless AI pilots get showcased in boardrooms yet generate almost no measurable financial return.
- First-mover advantage is compounding into an ever-widening performance gap: Leading companies learn faster, redeploy solutions faster, and automate decisions at a higher rate — advantages that stack into a self-reinforcing performance premium.
"AI Fitness": The Underlying Logic That Determines Winners
PwC defines "AI fitness" as the integrated capability to direct AI at what matters most, build fit-for-purpose foundational capabilities, and embed AI throughout the enterprise. Companies with higher AI fitness consistently outperform on the following intermediate performance indicators:
- Faster time-to-market for new products and services
- Stronger capacity to transform business models
- Improved decision quality
- Enhanced customer experience and trust
The Multiplier Effect of Foundational Capabilities
When companies strengthen their AI foundations while simultaneously expanding AI use, the improvement in AI-driven performance is nearly double that seen by companies with weaker foundations. This "2x conversion rate effect" points to a fundamental principle:
AI Use × AI Foundational Capability = Exponential, Not Linear, Performance Gains
A Three-Layer Framework for AI Use Cases: From Efficiency to Growth to Reinvention
PwC's research reveals a clear hierarchy of AI use case value — from low-value, isolated efficiency gains to high-value business model reinvention — forming a three-layer progressive structure:
Layer 3: Reinvention ← Highest value, hardest to achieve
↑ Cross-industry ecosystem collaboration, new business models, new value pool discovery
Layer 2: Growth ← High-to-medium value, where leading companies focus
↑ New products/services, new customer segments, new market entry
Layer 1: Efficiency ← Foundational value, where most companies remain stuck
↑ Process automation, cost reduction, speed improvement
Leading companies treat AI as a top-line-boosting reinvention engine, not merely an efficiency tool. They extend the technology's utility across all business transformation activities: from opportunity discovery, to creating new offerings, to reshaping their business models entirely.
A Full Spectrum of AI Use Cases
Efficiency Use Cases
This is the foundational layer of AI application — and the primary focus of the vast majority of enterprise AI projects today.
Intelligent Customer Service and Contact Centers
Typical scenario: AI-powered omnichannel contact centers combining predictive intent modeling, adaptive dialogue management, and real-time analytics to support seamless human-AI collaboration.
PwC case evidence: After deploying an AI-driven omnichannel contact center, a major technology company reduced the time customers spent on the phone by 25%, decreased call transfer rates by as much as 60%, lifted its Net Promoter Score (NPS) by 7%, and improved customer satisfaction by 10%.
The core value equation:
- Cost side: Reduced agent handling volume, lower labor costs
- Experience side: Higher first-contact resolution rates, shorter customer wait times
- Data side: Every interaction generates structured insights that continuously improve the service model
Extended use cases:
- Upgrading traditional IVR (Interactive Voice Response) systems to conversational AI
- Sentiment recognition systems that monitor customer emotion in real time and trigger escalation alerts
- Simultaneous multilingual service capabilities that transcend geographic service boundaries
Code Generation and Software Engineering Acceleration
Typical scenario: AI-assisted programming, code review, test generation, and documentation automation.
PwC case evidence: After deploying AI agents to support end-to-end software development, a global retail giant reduced its software development cycle time by up to 60%, cut production errors by 50%, and made significant inroads into a large IT backlog.
Key insight: AI application in software has a self-reinforcing quality — AI helps engineers build better AI systems faster, creating an accelerating flywheel of compounding capability.
Legacy System Modernization
Typical scenario: Using generative AI to reverse-engineer legacy code, extract embedded business logic, and produce structured requirements documentation for modernized systems.
PwC case evidence: Southwest Airlines partnered with PwC to apply GenAI to legacy system reverse engineering, cutting backlog creation time by 50% — from ten weeks to five — saving over 200 hours, and generating more than 600 requirements, 90% of which were accepted as high quality.
Strategic significance: Legacy modernization is the bottleneck blocking AI transformation for countless traditional enterprises. GenAI turns this high-risk, high-cost engineering challenge into an industrializable, repeatable process.
Finance and Supply Chain Automation
Typical scenario: AI-powered forecasting, financial reconciliation, analytics, and operational monitoring automation.
PwC case evidence: Electric vehicle manufacturer Lucid compressed its end-to-end forecasting cycle from weeks to under one minute, designed and began scaling 14 AI-driven use cases in just ten weeks, and has since expanded beyond finance into procurement and operations — including an AI executive concierge supporting decisions across more than US$1 billion in capital investments.
Extended use cases:
- Automated accounts payable/receivable processing
- Fraud detection and anomalous transaction identification
- Tax compliance and report generation
- Supply chain demand forecasting and inventory optimization
Enterprise Brand Standards Operations
Typical scenario: AI agents compress the brand standards change management process from weeks of manual effort into hours of automated workflow.
PwC case evidence: Wyndham Hotels deployed AI agents that cut brand standards review time by 94% — with AI reviews running 20 times faster than manual processes — saving 40 to 80 hours per review cycle while maintaining rigorous brand consistency.
Healthcare Data Integration and Clinical Analytics Acceleration
Typical scenario: Unifying fragmented, non-structured clinical data — pathology reports, biomarker data, treatment histories, social determinants of health — scattered across siloed systems into an AI-ready data foundation.
PwC case evidence: After building an AI-ready oncology data foundation in partnership with PwC and Google Cloud, a leading healthcare organization gave care teams 50% faster access to analytics, while creating more than US$50 million in new value potential through research acceleration and life sciences partnerships.
Growth Use Cases
This is the mid-to-high value layer of AI value creation, where leading companies are placing their strategic bets.
New Product and Service Innovation
Typical scenario: AI-assisted identification of unmet customer needs and design of entirely new products and services.
PwC case evidence: John Deere deployed its See & Spray AI precision spraying system, which covered more than one million acres during the 2024 growing season, saving farmers an estimated eight million gallons of herbicide mix with an average herbicide savings of 59% across corn, soybean, and cotton fields. More importantly, the technology was packaged into a commercial model allowing customers to pay for verified outcomes — opening a scalable services revenue stream for the company.
Strategic insight: The real breakthrough in this case is not the AI product innovation itself, but how AI enabled the transformation from one-time hardware differentiation to a recurring service revenue model. Business model reinvention is the true value multiplier.
AI-Assisted Personalization and Customer Experience Optimization
Typical scenario: Leveraging real-time behavioral data and predictive models to deliver personalized recommendations, pricing, and service to every customer.
Supporting data: After deploying AI agents, a major retail company reduced customer response times by up to 40%, increased marketing conversion rates by 15%, and improved marketing ROI by 20%.
Extended use cases:
- Real-time personalized product recommendations on e-commerce platforms
- Personalized financial planning recommendations from banking institutions
- Individualized preventive health programs in healthcare
AI-Driven Marketing and Sales Intelligence
Typical scenario: Using AI to scan market signals, identify high-potential customers, and optimize the marketing mix.
Industry distribution: Technology services and hospitality and leisure sectors show notably high AI adoption in demand generation functions such as marketing and sales.
Representative use cases:
- Predictive sales lead scoring
- AI-generated marketing content — ad copy, images, video
- Dynamic pricing optimization
- Intelligent sales conversation analysis to detect buying signals
Workforce Planning and Employee Retention
Typical scenario: Using AI to analyze employee behavioral data, predict attrition risk, and optimize scheduling and performance management.
PwC case evidence: A major retailer reduced employee attrition by 10% through AI-driven workforce planning.
Extended value: AI application in HR is expanding from initial talent acquisition screening to employee career path planning — fundamentally reshaping the logic of human capital management.
Business Model Reinvention Use Cases
This is the highest layer of AI value creation and the single most important dimension distinguishing leading companies from the rest.
Cross-Industry Value Pool Discovery and Ecosystem Collaboration
Leading companies are 1.8 times more likely than others to use AI to identify emerging value pools — particularly those centered on customer needs that call for innovative, multi-sector combinations of products and services. Capturing growth opportunities arising from industry convergence stands out in PwC's research as the single strongest AI fitness factor influencing AI-driven financial performance.
Landmark case: Automotive manufacturers and healthcare providers collaborating to equip vehicles with advanced health monitoring sensors, feeding that data to AI systems that then design personalized preventive health programs.
Additional cross-industry convergence scenarios (analytical projections):
| Industry A | Industry B | Convergence Scenario | AI's Role |
|---|---|---|---|
| Insurance | Healthcare | Behavioral insurance (dynamic premiums based on health data) | Risk model construction + data integration |
| Financial Services | Retail | Embedded finance (lending/insurance at point of purchase) | Real-time credit assessment + fraud detection |
| Automotive | Energy | Vehicle-to-Grid (V2G) | Intelligent charging schedule optimization |
| Agriculture | Life Sciences | Precision agriculture meets biotechnology | Crop gene-environment interaction prediction |
| Education | Labor Markets | Real-time skills-to-job matching | Personalized learning pathways + employment prediction |
AI-Driven Decision Automation Maturation
Among all the operational performance indicators PwC tested, decision automation has the strongest link to AI-driven financial performance. Leading companies make 2.8 times as many decisions without human intervention as their peers — and they report significantly stronger gains in decision quality, a reminder that automation, done well, improves quality alongside speed.
Decision Automation Maturity Model:
Level 1: Assistance
→ AI provides recommendations; humans decide
→ Examples: Credit approval suggestions, inventory replenishment recommendations
Level 2: Augmentation
→ AI handles routine decisions within guardrails; humans manage exceptions
→ Examples: Automated low-risk insurance claims processing, intelligent routing
Level 3: Automation
→ AI executes multi-step tasks within defined guardrails
→ Examples: End-to-end procurement workflows, compliance checks
Level 4: Autonomy
→ AI operates and self-improves independently
(only 15% of AI leaders have reached this level)
→ Examples: Quantitative trading strategies, adaptive cybersecurity response
Key constraint: Only 15% of AI-leading companies report that their most sophisticated use case has reached the autonomous and self-improving level. While 48% of AI leaders anticipate head-count reductions of at least 5% due to AI, 49% expect little to no change in workforce size — or even net increases.
Enterprise-Scale Agentic AI Networks
Typical scenario: Building an enterprise-wide AI agent hub — a centralized platform for prototyping, deploying, and governing AI agents at scale.
PwC case evidence: A global retail leader built a centralized AI hub as a universal platform to prototype, deploy, and govern AI agents. After the first wave of agents supported end-to-end software development, subsequent waves expanded into customer service and people management, forming a full enterprise AI agent ecosystem.
Strategic implication: The value of AI agents lies not in any single agent's capabilities, but in the collaborative network among agents. This represents a fundamental leap — from "point-solution AI deployments" to "systemic AI infrastructure."
AI Use Case Distribution Across Industries
PwC's research data reveals that AI embeddedness priorities vary meaningfully by sector.
Media and entertainment companies show notably high AI integration across the value chain: 54% in direction-setting functions (e.g., strategy and planning), 55% in demand generation (e.g., marketing and sales), 35% in support services (e.g., finance and HR), and 41% in demand fulfillment (e.g., production and supply chain planning). Pharmaceuticals, life sciences, and automotive lead in direction-setting; technology services and hospitality and leisure lead in demand generation; private equity leads in support services; and insurance leads in demand fulfillment.
An Industry-by-Industry AI Use Case Map
Financial Services (Banking / Insurance / Asset Management):
- Intelligent risk assessment and compliance automation
- AI-powered quantitative investing and portfolio management
- Insurance underwriting and claims processing automation
- Real-time anti-money laundering and fraud detection
- Personalized wealth management advisory
Healthcare and Life Sciences:
- AI-assisted medical imaging diagnostics
- Clinical trial patient matching and acceleration
- Drug molecule screening and target discovery
- Personalized treatment plan recommendations
- Hospital operations scheduling and optimization
Manufacturing and Industrial:
- Predictive maintenance (equipment failure prediction)
- Automated quality inspection using computer vision
- Precision agriculture (the See & Spray model)
- Supply chain resilience optimization
- Digital twin-assisted design
Retail and Consumer Goods:
- Personalized recommendation engines
- Dynamic pricing and promotion optimization
- Demand forecasting and inventory management
- Cashierless checkout and automated warehousing
- Social media sentiment analysis
Media and Technology:
- AI-generated content (text, images, video, music)
- Content recommendation algorithms
- Targeted advertising optimization
- Code generation and software engineering acceleration
- Platform safety and content moderation
AI Foundational Capabilities: The Bedrock of Use Case Success
Successful use cases do not exist in isolation — they rest on six foundational capabilities. These are among the most underappreciated findings in PwC's research, yet they are precisely what determines whether a use case can be replicated and scaled reliably.
Strategic Discipline and Investment Commitment
AI-leading companies invest 2.5 times as much of their revenue in AI as their peers. Leaders in software, banking, and media and entertainment report the highest investment levels — approximately 5% of annual revenue. But dollar commitment alone is not sufficient: what also sets leaders apart is dynamic resource reallocation. Leading companies are 1.3 times more likely to shift financial and human resources toward high-value AI projects as their business priorities evolve.
Innovation Infrastructure
AI leaders are 1.5 times more likely to simultaneously provide dedicated AI experimentation infrastructure (such as sandboxed environments isolated from enterprise systems) and appoint innovation owners within business units. This "dual-track model" — purpose-built technical infrastructure plus embedded business accountability — is the key combination that sustains high-velocity innovation.
Employee Trust and Adoption
Employees at AI-leading organizations are 2.1 times more likely to trust AI-generated insights and act on them in their day-to-day work. Building this trust requires three systemic elements:
- Involvement: Cross-functional teams co-create AI solutions alongside data and AI specialists, eliminating the clunky developer-to-user handoffs that kill adoption.
- Skill building: Ongoing, role-based AI learning that equips employees to apply AI in real work contexts.
- Safety guardrails: Clear boundaries around what AI is permitted to do, who is accountable, and what requires human escalation.
Responsible AI Governance
Leading companies are 1.7 times more likely to operate under a documented Responsible AI framework applied throughout the use case lifecycle, and 1.5 times more likely to have a cross-functional AI governance board. Effective governance is not a speed brake — it is an accelerant. It keeps routine use cases moving quickly, reserving board review only for the highest-risk decisions.
Data and Technology Infrastructure
Leading companies are 2.4 times more likely to create reusable, centrally catalogued AI components that teams can access off the shelf, and 1.7 times more likely to provide the high-quality data required for prioritized AI applications.
Key insight: A library of reusable AI components — including data pipelines and integration layers — is the core asset that drives down the marginal cost of each new AI deployment and unlocks true economies of scale.
AI Portfolio Management Discipline
Leading companies are more likely to conduct structured reviews of their AI initiative portfolios — to decide which to prioritize, scale, or terminate. Yet even among AI leaders, only 28% say they conduct such portfolio reviews to terminate initiatives to a large or very large extent, revealing that even the best-performing organizations still have meaningful room to sharpen their capital discipline.
Scaling AI Across the Enterprise: Three Dimensions of Transformation
Go Broad: Span the Value Chain
Leading companies are roughly twice as likely as others to have AI scaled or embedded across major parts of their value chain — from corporate strategy and supply chain operations to front-office customer engagement and back-office support functions.
Practical guidance: Choose one priority workflow and conduct an end-to-end review. Redesign the process around how AI will change handoffs, roles, and throughput — not just how to accelerate a single step within it.
Go Deep: Embed AI in Core Workflows
Leading companies do not bolt AI on top of existing workflows — they integrate AI deeply into standard operating processes. Consider the contrast:
| Shallow AI Integration | Deep AI Embedding |
|---|---|
| A standalone chatbot that agents must consult and then manually copy results into a support ticket | AI running inside the case management system, automatically pulling customer context, drafting responses, and routing only complex cases to specialists |
| AI as an auxiliary tool that leaves underlying processes unchanged | AI reshapes process design; roles and responsibilities evolve accordingly |
| Humans and AI working in parallel silos, limiting efficiency gains | True human-AI collaboration, with quality and throughput improving simultaneously |
Go Autonomous: Expand the Frontier of Automated Decision-Making
Among all operational performance indicators tested, decision automation shows the strongest link to AI-driven performance. Leading companies make 2.8 times as many decisions without human intervention and report substantially stronger gains in decision quality.
A practical scaling strategy for automated decision-making: Begin with decisions that are high-frequency, repeatable, measurable, and carry low-to-moderate risk — such as triage, prioritization, and routing. Automate within explicit guardrails, instrument decision quality continuously, and expand the scope of automation only when reliability and trust thresholds have been demonstrably met.
Extended Analysis: Strategic Insights Beyond the Report's Frame
The "Time Dimension" of AI Use Case Value Is Severely Underestimated
Most companies today evaluate AI use cases using static ROI models — a one-time benefit compared against upfront investment costs. But PwC's data reveals a dynamic reality: AI use case value compounds over time.
The compounding mechanism operates through three reinforcing loops:
- The data flywheel: AI use generates data → data improves models → better models create more value → more usage
- Capability accumulation: Foundational capabilities built for early use cases (data pipelines, reusable components) reduce the development cost of every subsequent use case
- Organizational learning: Rising employee trust in AI → higher adoption rates → richer data quality → improved model performance
The implication is profound: a "mediocre" AI project today may become a formidable competitive moat tomorrow. Conversely, companies that wait face an exponentially steeper hill to climb.
Industry Convergence: The Largest Untapped AI Value Pool
Leading companies are two to three times more likely than peers to use AI to collaborate with companies in other sectors to unlock value, to work within multi-industry business ecosystems, and to compete in markets beyond their traditional sector boundaries.
This finding carries a strategic imperative: the largest AI use cases of the future will not reside within any single industry — they will emerge at the boundaries where industries intersect. Enterprises need to build a standing "industry radar" — a systematic capability to continuously scan for opportunities where their core competencies meet unmet needs in adjacent sectors.
Illustrative untapped value pools:
- "Health-as-a-Service": Insurance × Healthcare × Wearables × AI → Personalized health management with dynamic premiums
- "Mobility-as-a-Service": Automakers × Urban Transit × Energy × AI → Subscription-based, seamlessly integrated mobility
- "Learn-to-Earn": Educational Institutions × Recruitment Platforms × Employers × AI → Skills credentialing with direct employment pathways
- "Farm-to-Fork": Agriculture × Food Safety × Retail × AI → Full-chain traceability and quality assurance
The Counter-Intuitive Value of AI Governance: Not a Brake, but an Accelerator
Most enterprises treat AI governance as a compliance burden. Yet PwC's data consistently shows that companies with stronger governance capabilities also tend to move faster in AI deployment.
The mechanism is straightforward:
- Standardized build templates eliminate repeated reinvention of foundational components
- Rapid checkpoint mechanisms allow routine use cases to advance without waiting for full committee review
- Clear accountability structures give teams the confidence to take calculated risks on higher-stakes initiatives
- The trust foundation built by a Responsible AI framework drives the employee adoption that determines whether AI delivers real-world impact
Conclusion: The ROI of governance investment extends far beyond risk mitigation — it is, above all, a speed dividend.
From "AI Projects" to "AI-Native Enterprise": The Deep Waters of Organizational Transformation
PwC's research surfaces a dimension that has yet to receive sufficient strategic attention: the deepest challenges in AI transformation are not technological — they are organizational and operational.
What leading companies are actually building is an entirely new "human-machine collaboration operating system":
- New roles: AI Orchestrators, AI Overseers, AI Optimizers
- New processes: Rules for task allocation between humans and agents, quality checkpoints, escalation protocols
- New incentive structures: Performance frameworks that reward AI experimentation and recognize employees who surface scalable AI solutions
The organizational depth this transformation demands far exceeds the technical complexity of any single AI use case.
Logical Audit: Potential Inconsistencies and Reader Advisories
Advisory 1: The Risk of Conflating Correlation with Causation
The report infers a roadmap for "becoming an AI leader" from the observed characteristics of AI-leading companies — an approach that carries a meaningful survivorship bias risk:
- Companies with high AI investment and mature governance structures may simply be organizations with stronger overall management discipline
- The concentration of AI returns may partly reflect differences in general organizational excellence, not purely differences in AI strategy
- Recommendation: Before applying the report's prescriptions, honestly assess your organization's baseline management maturity — and resist the temptation to simply copy AI best practices onto a weak organizational foundation
Advisory 2: The 7.2x Figure Lacks an Absolute Baseline
"The most AI-fit companies deliver AI-driven financial performance 7.2 times as high as other respondents" is a striking headline — but the report does not clearly disclose:
- What the absolute AI-driven financial performance of the baseline group (the "other companies") actually is
- Whether 7.2x represents a multiple of near-zero performance, or a multiple of already-significant performance
- Recommendation: Treat the 7.2x figure as directional evidence of the performance gap, not as a precise forecast for your own organization
Advisory 3: The Tension Between Industry Convergence Opportunities and Privacy Regulation Goes Underexplored
The report's "automotive × healthcare" sensor monitoring scenario is compelling — but the text does not adequately address:
- The legal and compliance challenges of cross-industry data sharing (GDPR, medical data protection regulations, etc.)
- The degree of consumer acceptance for the commercial use of personal health data
- The complexity of data sovereignty in cross-border operations
- Recommendation: Privacy regulation and data sovereignty must be treated as core design constraints in cross-industry AI strategy from day one — not as compliance considerations to address after the fact
Advisory 4: An Internal Tension Around the "28% Portfolio Review" Finding
The report simultaneously emphasizes the "portfolio management discipline" of leading companies and notes that even among AI leaders, only 28% conduct portfolio reviews to terminate initiatives to a large or very large extent. This suggests that even the highest-performing organizations have a meaningful gap in their AI project culling mechanisms — a finding that sits in some tension with the report's broader characterizations of AI leader discipline, and one that warrants careful reader attention.
Advisory 5: Sample Skew Toward Large Public Companies Limits Generalizability
91% of the research sample are publicly listed companies; 76% have annual revenues exceeding US$1 billion. This means the study's conclusions have limited direct applicability to small and medium-sized enterprises:
- An AI investment target of 5% of annual revenue may simply be unattainable for smaller organizations
- The cost of building dedicated AI sandbox environments and cross-functional governance boards can be prohibitive at smaller scale
- Meaningful participation in cross-industry ecosystem plays typically requires a degree of market position and negotiating leverage that smaller firms may lack
From "A Collection of AI Pilots" to "An AI-Driven Enterprise"
PwC's research delivers a clear and sobering signal: the distribution of AI value is polarizing — and the speed of that polarization will only accelerate as leading companies' compounding advantages widen the gap.
For business leaders, the core action framework can be distilled into three essential questions:
- Is our AI pointed at what truly matters? — Not just cost savings, but growth and reinvention
- Are our AI foundations strong enough to support repeatable scale? — Not project-by-project heroics, but a system that reliably converts AI investment into AI outcomes
- Is AI running broadly, deeply, and autonomously across the enterprise? — Not a portfolio of pilots, but an AI-native operating model
When AI is trusted, aimed at reinvention, supported by targeted foundational investments, and scaled through repeatable patterns across workflows and decisions, the results transcend incremental improvement — they add up to a compounding performance premium.
This article is based on PwC's research report "Want ROI from AI? Go for Growth" (published April 13, 2026), synthesized, analyzed, and extended with reference to publicly available data and analytical projections. Inferential conclusions do not represent the official position of PwC.