AI-driven workflow automation: what McKinsey revealed about the companies that are winning
Workflow automation is no longer a trend — it has become a matter of survival in the corporate world. Over the past three and a half years, since ChatGPT first landed on most people’s radar, the game has changed in ways that can’t be undone. Artificial intelligence left the lab, moved past the machine learning phase, and reached a new stage: agentic AI, capable of automating entire processes end to end without requiring human intervention at every step. And the companies that figured this out early are pulling ahead by a significant margin.
We’re not talking about a marginal edge here. We’re talking about a 20% average increase in EBITDA, a three-dollar return for every one dollar invested, and an execution speed that leaves competitors looking in the rearview mirror wondering what just happened. McKinsey just released the second edition of Rewired, its framework for technology transformation in large enterprises, and the numbers are hard to ignore. What changed, who’s getting it right in practice, and what separates the organizations harvesting real results from those falling behind — that’s exactly what we’re going to break down here. 🚀
Why McKinsey revisited Rewired now
The first edition of Rewired gave business leaders a clear structure for how organizations could empower themselves with technology and artificial intelligence. But the world didn’t leave much room for debate about the need for a second version, as Rob Levin, one of the people behind the project, explained. Technology moved too fast. AI evolved from machine learning models to agentic AI — this new ability to automate complete workflows from start to finish.
The team behind the framework wanted to revisit the recipe they had established three years earlier and ask an honest question: does the approach for traditional enterprises to organize, align, build, adopt, and scale AI still hold up? The answer came with a quote from the Greek poet Archilochus that captures the moment perfectly: we don’t rise to the level of our expectations, we fall to the level of our training. And that remains true both for the current state of AI and for the central thesis of Rewired.
The core insight is that most companies within a given industry agree on where AI can create value, which domains can be transformed, and what the big business cases are. But companies only perform at the level of their internal capabilities. And the big finding of this second edition was exactly that: companies that built strong capabilities during the early AI phase — the so-called AI 1.0 — had far greater success transitioning to generative and agentic AI than those that never invested in that foundation.
What makes agentic AI different
For years, when people talked about artificial intelligence in a business context, what they really had in mind was a very sophisticated assistant: you asked, it answered. The model worked fine for one-off tasks but had a clear ceiling. Whenever a process needed to move to the next step, a human had to be there to push the button. That created bottlenecks, delayed deliverables, and capped the real potential of the technology. The breakthrough agentic AI brings to the table is exactly the ability to break that cycle.
Agentic AI doesn’t wait to be called. It acts. In practical terms, an AI agent can receive an objective, plan the necessary steps, execute each one sequentially or in parallel, adapt the plan based on intermediate results, and deliver the final output without anyone needing to step in along the way. This radically changes the logic of corporate workflows, because the human bottleneck that always existed between process steps simply disappears — or at the very least, is drastically reduced.
This leap didn’t come out of nowhere. It’s the result of a combination of advances in large language models, improvements in contextual reasoning capabilities, and the arrival of orchestration infrastructure that allows multiple agents to work in concert. When you put it all together, the result is a level of automation that goes far beyond what any company could have imagined five years ago. And the most important part: this level is accessible right now — it’s not something reserved for the future.
The numbers McKinsey brought to the table and why they matter
The Rewired framework, in its second edition, is not a theoretical document packed with buzzwords. It’s a compilation of real data from large organizations that have already put AI-driven digital transformation into practice and measured the results with rigor. Kate Smaje, one of the study’s leaders, shared three numbers that paint a pretty clear picture — all based on a cohort of 20 companies that actually applied the framework in full. 📊
The first number: those 20 companies recorded an average 20% increase in EBITDA. To put that in perspective, this is a metric that large corporations chase for entire years through operational efficiency programs, cost-cutting, and restructurings. AI is delivering that in much shorter windows of time.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
The second number: on average, it takes one to two years for the investment to become cash-cumulative. And an interesting detail is that two-thirds of this cohort achieved it by focusing on three or fewer areas of transformation. In other words, these companies aren’t spreading AI across the entire organization like peanut butter. They’re being surgical about where they direct resources.
The third number: for every dollar invested, the average return is three dollars. This matters because it dismantles an argument that still circulates frequently in boardrooms — that the cost of implementation is too high to justify the project. When average ROI hits 3x, the cost argument loses steam very quickly.
What Rewired makes clear is that these results are not universal. They belong specifically to companies that applied the methodology in an integrated way. The rest, who tried isolated elements without building that organizational muscle, fell short of those numbers. The real risk isn’t in investing — it’s in not investing while competitors forge ahead.
Real-world cases: Freeport-McMoRan, DBS, and LATAM Airlines
Theory only earns credibility when it has concrete examples behind it. And Rewired brings a few that are worth knowing.
Freeport-McMoRan, a giant in the mining sector, is a standout case. In the first edition of the book, the company was already highlighted for building a digital twin of its entire copper concentrator, creating end-to-end efficiencies in the process of extracting metal from rock. The results were impressive. When generative AI came along, the company channeled that same capability into another area of the business: leaching — the final chemical process for extracting ore from rock — and generated a significant new layer of value. This is a perfect example of how capabilities built during the early AI phase continue to serve companies incredibly well as the technology evolves.
DBS, the Singapore-based bank, is another case the study revisits with emphasis. Over three to four years, DBS made heavy investments in technology foundations. When generative and agentic AI arrived, the bank was able to move with impressive speed. The tangible, verifiable result was roughly one billion Singapore dollars in AI-generated benefits. But the number that might be even more revealing is the speed metric: when DBS first started working with AI, it took 18 months to put the first model into production. Today, the bank puts a new model into production every two months. That acceleration is the differentiated capability that allows them to do more and move faster on an ongoing basis.
LATAM Airlines shows up as a reference in agentic engineering. According to the study, the airline is probably a year ahead of most companies in terms of adopting and embedding agentic engineering — not just for coding, but for the full software development lifecycle. And the resulting speed has been remarkable. 🛫
Where workflow automation is generating the most impact
When we look at the use cases that stand out the most in practice, certain sectors and functions come up more often. Financial operations, customer service, supply chain, content generation, and data analysis are the areas where AI-based automation has delivered the most impressive results. And the pattern that keeps repeating across these cases isn’t total replacement of teams — it’s the transformation of the role people play within the process. Teams shift to working on exceptions, strategic decisions, and continuous improvements, while AI handles the operational volume. 🤖
In retail, for example, forecasting and planning is where AI tends to generate the greatest impact. In insurance, claims processing is the most common target. In heavy manufacturing, the focus tends to be yield or throughput. This specificity is critical, because one of the most common traps is trying to apply AI to everything at once — diluting resources and making it impossible to measure real impact in any single area.
In operations and supply chain, the impact is equally striking. Workflows that used to involve multiple manual approvals, email exchanges between departments, and days of waiting are now executed in minutes. An AI agent can monitor inventory levels, identify stockout risks, query alternative suppliers, compare prices, and trigger orders automatically — all within a single orchestrated flow. That kind of operational capability was unthinkable without a dedicated, expensive team. Today, it’s infrastructure.
One point raised in the study that deserves special attention is the case of a major automotive company that completely reinvented its supply chain. As difficult as the technological transformation itself was, the work that followed was just as hard: convincing hundreds of suppliers to operate the way the supply chain’s digital twin indicated was optimal. This kind of adoption challenge — one that goes well beyond the technical side — is frequently treated as a minor detail and ends up becoming the main obstacle to success.
The quiet revolution in software development
One of the most impactful points Rewired addresses is the radical shift in software development productivity. We’re talking about up to a 20x increase in productivity for code creation, driven by AI tools. This is not a future projection. It’s already happening.
This shift is collapsing the traditional model for development teams. The two-pizza team concept, popularized by tech companies like Amazon — which assumes roughly eight people working together on a project — is being reduced to two professionals: a product owner who knows how to define what needs to be done and what the ideal outcome looks like, and a full-stack engineer who can work with code generation systems, debug results, and integrate everything into the existing architecture. This is a massive change in work dynamics and in the type of talent organizations need to pursue. ⚙️
At the same time, the technology landscape has never been more complex for making architecture decisions. Every vendor at every layer of the stack is positioning itself as the center of gravity for your AI strategy, and many of these vendors come with steep ongoing operational costs. The temptation to adopt point solutions or agent platforms for every business function is real, but when you step back and look at the big picture, the result can be a fragile, inefficient, and less secure architecture.
What sets the winners apart from those stuck in pilot mode
There’s a very common phenomenon in large enterprises that’s being called the eternal pilot: the organization runs an AI project, gets promising results at a small scale, and then stalls. The project doesn’t scale, doesn’t integrate into the real business, and turns into an internal showcase that generates no measurable financial result. This happens for very specific reasons, and understanding them is the first step toward avoiding the trap.
The main reason is a lack of integration between business strategy and AI strategy. When an automation project is born inside an isolated department, with no connection to the company’s financial objectives and no real executive sponsorship, it rarely makes it past the demo phase. True transformation happens when company leadership sees workflow automation as a strategic lever — not as a tech experiment.
One of the first mistakes identified by the study is the mindset that leadership’s role is to hear proposals about AI, allocate resources, and then delegate execution to the technology team. This approach replicates an outdated paradigm of working with IT that simply doesn’t work anymore. Transformation needs to be led by the business and distributed across the entire leadership team.
Another decisive factor is building internal capabilities. Companies that depend entirely on outside vendors to run their AI systems become hostage to timelines, costs, and customization limitations. Those that invest in forming internal teams with real knowledge of how agents work, how workflows are orchestrated, and how models can be tuned for the business context have a huge advantage in iteration speed and long-term cost.
Who should lead this transformation
An inevitable question comes up when all these points converge: who owns this transformation? Where does final accountability sit? The answer, according to the study, is that it needs to be both top-down and distributed. Across the hundreds of transformations studied, none succeeded without the project being among the CEO’s top one or two priorities.
At the same time, execution needs to be distributed across the entire leadership team. Kate Smaje describes it as a corporate team sport. A telling sign of trouble, she says, is when someone asks a question in a management meeting and everyone in the room turns to look at the person who has AI in their job title. When that happens, it’s a sign things aren’t going to work out.
The CHRO needs to wake up every morning thinking about what an agentic organization looks like. The CFO needs to reconfigure funding mechanisms to enable continuous reinvestment. The business unit owners — who are the real heart of the transformation — need to take ownership as well. The era when it was possible to hand this off to the technology function and hope for the best is definitively over.
AI transformation is, first and foremost, a people transformation
Every AI transformation, at its core, is a people transformation. And that has never been more true than right now. Talent density in teams matters — especially technical talent. At the same time, as agents take on more coordination, routine tasks, execution, and decision-making, the human role moves up the value chain.
The study suggests that companies start thinking about what it means to have carbon and silicon employees working together in the same organization. That involves rethinking roles, training teams to work alongside agents, and accepting that the definition of many jobs is going to change fundamentally. If you’re not thinking about your AI transformation as a set of people changes, there’s probably something off about your approach. 👥
- Strategic integration: automation needs to be connected to the company’s financial and operational goals — not just to the IT roadmap.
- Real executive sponsorship: transformation projects without backing from senior leadership rarely scale beyond the pilot stage.
- Skilled internal teams: relying solely on third parties to run AI limits the speed of evolution and increases long-term costs.
- Focus on high-impact workflows: starting with the processes that have the highest volume, highest cost, or greatest risk of error is the fastest path to measurable results.
- Think in domains, not use cases: three or fewer focus areas that move the needle are more effective than dozens of scattered projects.
- Clear metrics from the start: without indicators defined before implementation, it’s impossible to know whether the project is delivering real value.
- Plan for adoption and scale: planning only through the MVP isn’t enough. You need to think about production, repeatability, and scale from day one.
Building conviction around AI
One of the final points addressed in the study touches on something that doesn’t show up on any spreadsheet: conviction. How do you build conviction about AI — both for yourself as a leader and for the entire organization?
The best way to build that conviction is to follow the value. Go where the money is and solve real business problems. That’s the heart of it. It’s not about adopting technology because it’s new or impressive — it’s about proving that it solves concrete problems that matter to the company’s bottom line.
And then, a piece of advice that might seem counterintuitive: give yourself a little grace. This is hard, and one of the benefits of agentic AI is that the cost of iteration has dropped. It’s easier to get it wrong, course-correct, and pivot. The process is becoming less about finding the perfect answer on the first try and more about testing, validating, and building conviction along the way. Sometimes, what ends up on the cutting room floor — and the reason it was discarded — is where the real value hides. 💡
The time is now
The moment we’re living through is rare. Technology has advanced fast enough that workflow automation with artificial intelligence is viable at real scale, but there’s still a window of competitive advantage for companies that move first. That window won’t stay open forever. As tools become more accessible and use cases more widely known, differentiation will increasingly depend on the quality of execution and the depth of capabilities built internally.
The companies standing out right now aren’t necessarily the biggest or the wealthiest. They’re the ones that built the organizational muscle to turn insight into decision and decision into action with speed. They operate at a different metabolic rate, as Kate Smaje described it. The latency between spotting an opportunity and capturing it is fundamentally shorter. And that speed isn’t just operational — it’s cumulative. Each learning cycle makes the next one faster and more efficient.
Those who start now will arrive at the market’s maturity point with more experience, more data, and more advantage. And those who wait will face the reality that their competitors are already operating on a different level. AI-driven transformation is no longer one option among many. It’s the path forward.
