The takeaways
- The most AI fit companies in our research deliver AI-driven revenues and efficiencies that are 7.2x as high as those of other companies.
- When companies strengthen their AI foundations—strategy, investment, data and technology, workforce, governance and risk, and innovation—they generate far greater performance improvement as they increase their AI use.
- Leading companies treat AI like a reinvention engine, using it to reshape business models and pursue growth opportunities as industries converge.
In conference rooms from New York to Singapore, the same moment occurs again and again. Someone pulls up a slide with a tidy grid of AI pilots—chatbots here, decision engines there—and the room nods along. Then the questions begin. Which of these pilots are increasing revenue? Which are driving costs down? How many decisions have been made better, faster, safer?
The silence that often follows reflects an uncomfortable reality: for many companies, all that AI activity isn’t producing measurable returns. PwC research finds that value is currently concentrated in a small cohort: 20% of the 1,217 companies we surveyed capture 74% of the AI-driven returns.
What separates these AI leaders from the rest? It’s what we’ve come to define as “AI fitness”: the ability to point artificial intelligence at what matters, build fit-for-purpose foundations, and embed AI throughout the enterprise.
This article is written for business leaders who want to stop counting AI pilots and start driving measurable revenue gains and cost savings with AI. It explains what companies that are seeing outsized results do to become AI fit—and their practices are well within reach of all businesses.
The 7.2x performance advantage
To understand why some companies are seeing real returns while most are not, we benchmarked 1,217 companies—spanning regions around the world and 25 sectors—on their AI-driven financial performance, defined as the revenue and efficiency gains derived from AI and adjusted so each company could be compared against its sector's median.
We also asked senior executives at these companies about their engagement in 60 areas of AI management and investment practice to test those areas’ effects on AI-driven financial performance. We grouped these practices into nine categories related to the ways in which companies use AI and the foundational capabilities that make AI reliable and scalable, such as strategy and governance. These nine categories make up the components of our AI fitness index.
The headline result is clear: the most AI fit companies in our research deliver AI-driven financial performance that’s 7.2 times as high as the other respondents’ performance.
The reason for this dominance? Higher AI fitness levels improve a broad set of intermediate performance outcomes that, in turn, shape financial results. Companies with the strongest AI-driven financial performance (our “AI leaders”) are more likely than others to report that their AI portfolio has sped up the launch of new products and services. They also say it has helped transform their business and operating models, improved the quality of their decision-making, and enhanced customer experience and trust—metrics many executives already focus on improving.
Idea in motion
A large technology provider improves customer experience
The prompt
A major technology company with millions of customers faced rising expectations for seamless, personalised service. But its largely manual customer engagement model couldn’t keep up. Company leaders wanted to improve customer experience while keeping costs under control.
The move
PwC designed and deployed an AI-driven, omnichannel contact centre that combined predictive intent modelling, adaptive dialogue, and real-time analytics to support humans and AI agents. A centralised AI agent management hub enabled orchestration across channels, scaled deployment, and governance. To help employees use the new software effectively, the company also stood up Responsible AI, workforce upskilling, and new ways of working for human–AI teams.
The outcome
The results were immediate and measurable: customers spent 25% less time on the phone to get requests resolved, and call transfers fell by as much as 60%, meaning more issues were handled on the first contact. Customer experience improved as well; the company’s Net Promoter Score (NPS) rose 7%, and customer satisfaction rose 10%.
Similar compounding effects occur between AI use and AI foundations. When companies with strong foundations increase AI use, they see nearly double the improvement in AI-driven performance seen by those that have weaker foundations. In effect, foundations raise the conversion rate from AI activity to measurable outcomes. Stronger data and platforms reduce time-to-deploy, while workflow redesign and workforce trust-building increase adoption. Greater adoption, in turn, generates richer data and feedback—improving the system over time and increasing impact with each deployment.
It’s clear that companies pulling ahead through AI aren’t simply “doing more AI.” They’re building the capabilities that make AI scalable and reliable and then choosing where to apply that scale for maximum financial leverage.
INSIGHT FOR ACTION
2x
improvement in AI-driven performance for companies that back up increased AI use with stronger foundations.
Why it matters
Delivering use cases without the ability to repeat them reliably delivers lower ROI.
Your next move
Before expanding your AI footprint, identify the one or two foundation capabilities most likely to block repeatability and fix them for the highest-value initiatives first.
What do the leading companies point AI at first? Not just incremental efficiency, but also reinvention and growth, particularly where value is moving as industries converge in a “value in motion” world.
Aim AI at growth and reinvention
Plenty of companies use AI to become more efficient at the work they already do. Think of insurance firms, where AI solutions rapidly process claims, or software makers, where programmers direct AI to write a substantial portion of new code. The AI leaders we studied use AI for efficiency, too. But they don’t stop there. These companies treat AI like a top-line-boosting reinvention engine—one that helps them create fresh offerings and reshape their business models to move into promising new markets.
Our study shows that leading companies are 2.6 times as likely as others to report that AI has improved their ability to reinvent their business model.
At leading companies, the technology’s utility spans all the business reinvention activities we studied. It starts with the search for opportunity. AI leaders, we found, are 1.8 times as likely as other companies to use AI to spot emerging value pools—in particular, value pools centred on customer needs that call for innovative, multi-sector combinations of products and services. As industries converge to fulfil these needs, the rewards accruing to companies that reinvent their business models will increase.
In fact, the ability to capture growth opportunities resulting from industry convergence stands out in our research as the single strongest AI fitness factor influencing AI-driven financial performance. AI leaders are two to three times as likely as others to use AI to collaborate with companies in other sectors, to unlock value by working in ecosystems of businesses, and to compete beyond their usual sectors. Consider the possibility of car manufacturers and healthcare providers working together to equip vehicles with high-tech sensors that monitor the driver’s health and feed the data to AI systems that then design personalised prevention programmes.
INSIGHT FOR ACTION
2.6x
as likely to say AI has helped you reinvent your business model—AI leaders versus the rest.
Why it matters
The biggest returns come when AI changes what you sell and how you create value, not just how fast you execute tasks.
Your next move
Identify two growth bets AI could unlock this year and define what proof that it works looks like.
The leading companies we studied also reinforce their AI-informed growth agendas with disciplined management. They make strategic choices early, and they operationalise those choices with ownership and measurement. Compared with others, leading companies are more likely to have a prioritised AI road map across near- and long-term horizons, to align AI vision with business objectives, to systematically track business impact, and to hold senior leaders directly accountable for AI outcomes.
INSIGHT FOR ACTION
80%
more likely to systematically track the business impact of AI initiatives if you’re a leading company versus part of the chasing pack.
Why it matters
There’s no way to know if your AI investments are delivering returns without a way to measure results.
Your next move
Stand up a monthly “scale or stop” review. Only projects with measured movement on a defined business metric get more funding.
Your next move: Shift from cost to cash. Treat “growth from industry convergence” as a distinct AI portfolio with senior sponsorship. Use AI to scan for where value is moving, then back that view with decisions: a prioritised road map, explicit owners, and impact metrics that force trade-offs.
Idea in motion
John Deere reinvents itself by adding AI
The
prompt
For farmers, rising input costs and sustainability pressures place greater importance on outcomes like reduced chemical use, higher yields, and better stewardship. For John Deere, these shifts mean opportunities to create value with innovative offerings that bring AI into more sophisticated machines. In response, John Deere has made it a priority to create a solutions-and-services business model that lowers upfront barriers and supports recurring, outcomes-linked revenue.
The
move
John Deere deployed See & Spray, an AI-powered ‘sense-and-act’ precision spraying system that uses boom-mounted cameras and onboard computing to identify weeds and trigger nozzles to squirt herbicides only where they’re needed. John Deere packaged the capability in a service-like commercial model that allowed customers to pay for verified outcomes.
The
outcome
In the 2024 growing season, John Deere reported that See & Spray was used on more than 1 million acres, saving farmers an estimated 8 million gallons of herbicide mix, with 59% average herbicide savings across corn, soybean, and cotton fields. Beyond offering these cost and sustainability gains for farmers, the model positions John Deere to capture more value from a scalable services revenue stream rather than a one-time hardware differentiator.
Build focused AI foundations
Aiming AI at reinvention and industry convergence opportunities is the easy part. The hard part is delivering those outcomes repeatedly—which is why the next differentiator is not ambition, but the six targeted foundations. Rather than treating foundations as an abstract modernisation agenda, AI leaders build only what’s needed to turn AI use that’s aimed at growth and other high-value business objectives into measurable outcomes at scale.
Foundations change the economics of AI. They reduce friction, rework, and ‘one-off’ builds, so each new deployment gets faster, cheaper, and more reliable.
As mentioned earlier, this shows up as a conversion-rate effect: after an underperforming company institutes the right practices, it should see double the payoff from new AI use cases.
Our research shows that the practices described below are the ones that lead to the greatest performance gains.
Fund and flex the AI portfolio like an investor
The leading companies in our study invest materially more in AI than other companies do: 2.5 times as much of their revenue. Leaders in the software, banking, and media and entertainment sectors report investing the most, about 5% of annual revenue. Ample investment in AI, however, is just part of the leaders’ formula. These companies also endeavour to keep their investments aligned with their business needs. According to our research, they’re 1.3 times as likely as other companies to reallocate financial and human resources towards high-value AI projects as their business priorities shift. That approach is consistent with a large body of research linking dynamic resource allocation to superior financial outcomes.
Foster AI-powered innovation
If funding is the fuel, innovation is the engine. AI leaders create the conditions for high-velocity experimentation. They’re 1.5 times as likely as other companies to provide tech infrastructure specifically to support AI experimentation: think of ‘sandbox’ environments, walled off from enterprise systems, where developers can safely try new AI solutions. These leaders are also more likely to appoint innovation owners who direct AI projects within business units. That combination makes it easier to kick off pilots and run them quickly and safely.
Moreover, AI leaders are more likely than others to carry out structured reviews of AI innovation efforts so they can decide which ones to prioritise, scale, or terminate. The result is a pipeline of experiments that reliably lead to value-generating AI solutions.
INSIGHT FOR ACTION
1.5x
AI leaders are 1.5x as likely to both provide dedicated infrastructure and support for AI innovation via designated owners in business units.
Why it matters
AI stalls when it’s everyone’s side job. Dedicated infrastructure plus accountable owners turns experimentation into repeatable delivery.
Your next move
Assign a named business owner and success metrics for each priority AI initiative with decision rights and pair them with a dedicated platform for experimentation and delivery.
To boost adoption, cultivate employees’ trust
AI value materialises when people use AI. That makes employees’ trust of the technology much more than a ‘change management’ line item. Lack of trust is a throughput constraint. Low trust means low use, which means low impact.
Leaders create the conditions for uptake. Employees at AI-leading organisations are 2.1 times as likely to trust AI-generated insights and act on them in day-to-day work.
What drives trust is rarely a single programme. It is a system composed of the following:
- Involvement. At leading companies, it’s more likely that teams of business, data, and AI specialists are co-creating AI solutions. Cross-functional collaboration promotes alignment between business needs and solution designs, and it does away with the clunky developer-to-user handoffs that kill adoption. AI leaders also offer employees clear incentives to experiment with AI, which can give rise to ideas for scalable solutions that support the firm’s priorities.
- Skill building. Effective upskilling teaches employees how to apply AI in everyday work situations. Leading companies are more likely to provide employees with this sort of ongoing, role-based AI learning. Their senior executives are also more likely to set good examples, by attending training sessions and visibly using AI in their work.
- Safety. AI leaders further build employee trust with guard rails and protocols. When people understand what AI is allowed to do, which matters require escalation, and who’s accountable, they can use AI with greater confidence. At leading companies, employees are more likely to have role-based controls on data and AI access, along with robust, up-to-date security for data, models, and infrastructure.
Use governance to further strengthen trust—and accelerate
Leading companies take governance seriously while applying it in a way that speeds up delivery rather than slowing it down. A governance board sets Responsible AI policies, and teams apply them in their day-to-day work through mechanisms like standard build templates, quick checkpoints, and regular monitoring. This keeps routine use cases moving quickly, as teams tap the board to review only the highest-risk work.
Companies that lead in AI are more likely to have this machinery in place: they’re 1.7 times as likely to use a documented Responsible AI framework that applies to processes from use case selection through application monitoring, and 1.5 times as likely to have a cross-functional AI governance board.
Idea in motion
Wyndham scales trusted AI agents
The
prompt
For Wyndham, a global hotel franchise, delivering a distinctive travel experience involves giving hotel owners the trustworthy, timely support they need to apply the company’s brand standards accurately yet have room for regional customisations. However, the process for changing brand standards averaged about 30 days of manual effort. Company leaders sought to improve this process. They put Responsible AI at the heart of their strategy to ensure a sound solution that employees felt confident adopting.
The
move
PwC helped Wyndham put trusted AI to work by designing agentic workflows with human oversight built in—using automated prompts, co-authoring, and real-time monitoring so teams could guide and oversee the agents. Wyndham also positioned the programme to scale with a Responsible AI framework and ongoing upskilling to build trust and adoption.
The
outcome
The agents consolidated standards, simplified workflows for change requests, and created centralised, user-friendly access for franchisees. Wyndham achieved brand consistency at speed without sacrificing rigour and reliability: review time for changes to brand standards dropped 94% (AI reviews were 20x as fast), saving 40–80 hours per review and positioning Wyndham to confidently apply trusted AI solutions across its operations.
Eliminate tech and data friction
In our experience, some of the biggest blockers to scaling AI are data quality and access, tech integration, and the hidden cost of rebuilding the same components (such as data pipelines and integration layers) repeatedly.
AI leaders focus on removing those bottlenecks for their high-stakes use cases.
They’re 2.4 times as likely to create reusable, centrally catalogued AI components that teams can pull off the shelf instead of reinventing. They’re also 1.7 times as likely to provide the high-quality data needed for prioritised AI applications.
Idea in motion
Southwest Airlines uses GenAI to turn legacy code into modernisation-ready requirements
The prompt
Southwest Airlines’ crew attendance and leave application ran on a legacy tech stack with limited documentation and heavy reliance on tacit knowledge. Executives resolved to find ways to make the system easier to maintain and upgrade—while managing the time, cost, and risk of modernisation.
The move
Southwest worked with PwC to apply GenAI and advanced software engineering to reverse-engineer the application's source code into clear functional requirements for the updated system and a prioritised modernisation backlog. Southwest knowledge specialists then validated and refined the outputs through workshops, producing a detailed delivery plan with greater confidence and a repeatable approach for future modernisation efforts.
The outcome
GenAI cut the time needed to create backlogs by 50%—from ten weeks to five—and saved more than 200 hours across engineering, technology, and business teams during planning and design. The work also produced upwards of 600 requirements, 90% of which were accepted as high-quality, reducing the risk of the modernisation effort before development began.
Your next move: Build only what your AI strategy demands rather than getting lost in an unending, broad-stroke transformation. That means anchoring foundations to a small set of priority outcomes, funding the portfolio in order to scale winners, modernising only the necessary data and platforms, and providing targeted workforce reskilling and governance. This message applies to both AI laggards and leaders—even the outperformers aren’t engaging in some best practices, which means they’re still leaving value on the table. For example, while AI leaders are more disciplined than peers about pruning initiatives, only 28% say they conduct AI portfolio reviews to terminate initiatives to a “large” or “very large” extent.
Idea in motion
A leading healthcare company turned data into actionable insight—and revenue potential
The
prompt
An industry-leading healthcare organisation knew its oncology data could help it deliver better care and accelerated research. But much of that information was trapped in siloed systems and unstructured notes. Even after the company modernised some of its platforms, key information like pathology, biomarkers, treatment history, and social determinants remained scattered. Executives resolved to unify this data so they could facilitate timely analysis and enable doctors to personalise care or match patients to trials.
The
move
With PwC and Google Cloud, the organisation built a scalable, AI-ready oncology data foundation that streamlined how data was ingested, cleaned, organised, and made searchable—across records, claims, third-party sources, and clinical notes. AI helped convert unstructured information into usable formats, while Google Cloud tooling delivered real-time insights designed around frontline clinical and research workflows, with embedded monitoring of data quality to build trust.
The
outcome
The programme organised about 2,000 data tables into reusable assets built for real-world decisions, such as recognising when a patient could benefit from more affordable—but equally effective—treatment options. Care teams now access analytics 50% faster, enabling quicker matching of patients to trials, point-of-care treatment comparisons, and earlier identification of risks. The privacy-protected insights also created more than US$50 million in new value potential through research acceleration and life sciences partnerships.
Embed AI across the enterprise
In the leading companies, once executives define the business objectives they hope to achieve with AI—growth, reinvention, efficiency, or some combination of those—they make sure AI solutions are developed and implemented everywhere in the enterprise that they can make a difference. Embedding AI across the enterprise involves working along three dimensions: implementing AI broadly across many parts of the business; infusing AI into core workflows and systems so it can enhance the execution of tasks; and applying AI in sophisticated ways, moving from assistance to automation.
Go broad
Our research indicates that most companies still concentrate AI in pockets, consisting of a few use cases scattered across a few functions. Leading companies scale proven use cases across teams, regions, functions, value chain activities, and products so that value is not trapped in one isolated area. For example, an insurer that proves AI can cut invoice processing time in finance can reuse the same document intake and workflow model to automate contract review in the legal function and claims processing in operations.
We found that AI leaders are roughly twice as likely as other companies to apply AI across the value chain, in areas as varied as corporate strategy, supply chain operations, and the front and back office.
INSIGHT FOR ACTION
2x
as likely as other companies to have AI scaled or embedded into major parts of the value chain if you’re part of the AI leaders group.
Why it matters
The biggest performance gains accrue when AI is consistently used where decisions get made and work gets done—and where processes are redesigned to maximise AI’s value.
Your next move
Choose one priority workflow and review it end to end. Redesign the process around where AI will change handoffs, roles, and throughput—not just speed up one step.
Some sectors are further along in using AI across the enterprise than others. Media and entertainment companies rate near the top for embedding AI into processes throughout the value chain, with 54% having done so in direction setting (e.g. strategy, planning), 55% in demand generation (e.g. marketing, sales), 35% in support services (e.g. finance, HR), and 41% in demand fulfilment (e.g. production, supply chain planning).
Other sectors rate well in particular parts of the value chain: direction setting for pharmaceuticals, life sciences, and automotive; demand generation for technology services and hospitality and leisure; support services for private equity; and demand fulfilment for insurance.
Go deep
The top-performing companies in our study don’t just add AI on top of workflows. They fully integrate AI into standard operating processes. That’s essential to improving both task efficiency and output quality. This could look like redesigning customer support so AI runs inside the case management system—pulling the right customer context and knowledge, drafting responses, and routing only complex cases to specialists—rather than bolting on a separate chatbot that agents have to consult and then manually copy back into a support ticket.
Idea in motion
Lucid goes finance first, then AI everywhere
The
prompt
As automaker Lucid prepared for its next phase of growth, executives wanted the finance department to evolve from reporting results to shaping them—improving the speed and quality of forecasting, planning, and decision support so finance could serve as a foundation for enterprise intelligence.
The
move
Working with PwC, Lucid rapidly prototyped AI-enabled forecasting and reporting capabilities using operational data, applied AI models, and agent-based tools. Cross-functional pods combined Lucid and PwC specialists to embed AI into finance workflows—automating forecasting, reconciliation, analytics, and monitoring, and creating a repeatable blueprint for scaling AI decision support across the business.
The
outcome
Lucid reduced end-to-end forecasting cycle time from weeks to less than a minute, and in ten weeks, designed and began scaling 14 AI-driven use cases. The work is now expanding beyond finance into such areas as procurement and operations, including an AI-enabled executive concierge that supports faster leadership decision-making with visibility into more than US$1 billion in capital investments.
Go autonomous
Across all the operational performance indicators we tested, automating decisions has one of the strongest links to AI-driven performance. The reason is simple: when AI can safely take on a larger share of routine, high-frequency decisions, cycle times shrink, throughput rises, and performance improvements emerge. Our research shows that AI-driven performance leaders are nearly twice as likely to operate AI at higher sophistication levels, meaning that AI executes multiple tasks within guard rails or even operates autonomously and self-improves. Perhaps it’s no surprise then that AI leaders are 2.8 times as likely to increase the number of decisions made without human intervention. These leaders also report much stronger gains in decision quality, a reminder that automation works best when quality improves alongside speed.
This doesn’t automatically mean machines are taking everyone’s jobs. Full autonomy is still the exception: only 15% of AI leaders say their most sophisticated use case is autonomous and self-improving. Plus, although 48% of AI leaders expect head-count reductions of at least 5% due to AI, another 49% expect either little to no change in head count, or head-count increases. Finally, in many cases, we’ve seen that the immediate shift is not the removal of people, but the removal of delay: AI handles repeatable judgment calls inside guard rails, while humans focus on exceptions, trade-offs, and the steering of decisions towards strategic objectives.
Your next move: Scale selectively. Pick a handful of priority use cases tied to your objectives, then industrialise them. This means redesigning the workflow from end to end to embed AI into processes and then replicating the pattern across teams, regions, functions, and decision points. A practical starting point to increase automated decision-making: begin with a small set of decisions that are high frequency, repeatable, and measurable, and that have low to moderate risk (for example, triage, prioritisation, and routing). Automate within explicit guard rails, instrument decision quality, and expand only when reliability and trust thresholds are met.
Idea in motion
A retail giant scales AI agents across the enterprise
The
prompt
Faced with growing pressure from nimble AI-native competitors, executives at a global retail leader knew they would need AI to drive productivity and business reinvention at enterprise scale—along with new ways of working, new processes, and an operating model that could manage risk while moving fast.
The
move
The company collaborated with PwC to build a centralised AI hub: a universal platform to prototype, deploy, and govern AI agents. The first wave of agents supported software development from end to end. Subsequent waves supported functions such as customer service and people management. In parallel, the company began reorganising for human–agent collaboration by upskilling talent, defining new roles, building trust through validation and ethics oversight, and establishing agent life-cycle management.
The
outcome
Within months, software development cycle times were as much as 60% shorter, and production errors had fallen 50%, which helped teams attack a large IT backlog. As the company introduced agents in more functions, customer response times dropped by as much as 40%; attrition fell 10% through improved workforce planning; fraud declined 25% via real-time transaction monitoring; and marketing performance improved, with 15% higher conversions and 20% higher ROI.
AI is more than capable of producing quantifiable benefits. However, the opening scenario of this article—rooms full of AI pilots and too little measurable impact—will keep repeating for companies that don’t work on building AI fitness. Our research offers a clear and encouraging path to measurable gains. What separates AI leaders is the set of management choices they make: aligning AI uses to critical business outcomes, building fit-for-purpose foundations, and embedding AI across the enterprise.
Putting that formula into place requires deliberate, sustained effort. It won’t be easy, not with the myriad priorities calling for executives’ attention. Still, companies that want to catch up to the leaders can’t afford to wait. The advantage that the AI leaders already enjoy will only grow, because these companies are learning fast, redeploying solutions faster, and safely automating decisions.
The time has come to think beyond pilots and aim higher. Executives should look to point AI towards the biggest strategic moves on the table and establish an operating model that turns AI investments into AI-driven performance. When AI is trusted, aimed at reinvention, supported by targeted foundations, and scaled through repeatable patterns across workflows and decisions, the results go beyond incremental improvement—they add up to a compounding performance premium.
About the authors
