Data, Analytics & AI

Getting AI Right: Focus on What Matters

| Charlie Ranalli

The article I shared in October provoked some interesting conversations. Customers, colleagues, and executives came with questions, pushback, and — more often than I expected — relief. Relief that someone was saying out loud what many were sensing: AI is real, but the timeline to impact is longer than the noise suggests.

Impact continues to trail hoopla

The loudest voices in AI right now belong to the people building and funding it. Altman, Huang, Pichai — their message is consistent and urgent: transformation is imminent, disruption is sweeping, the window to act is closing fast. This is not analysis. It is positioning. Their valuations, their funding rounds, their market narratives all depend on the world believing that AI will reshape everything, soon. They have every structural incentive to compress the timeline.

The disciplined research tells a different story. McKinsey’s latest global survey finds that more than 80 percent of companies are not yet seeing bottom-line impact from their AI investments — despite widespread experimentation. A cross-country study by the Bank of England and research partners surveyed nearly 6,000 executives across the US, UK, Germany, and Australia: over the past three years, AI has had “essentially zero” impact on employment, with more than 90 percent of firms reporting no measurable effect. The Economist’s analysis shows that tech employment has stagnated since 2022, but the cause is post-pandemic correction and interest rate cycles — not AI displacement. Oxford Economics found that AI was cited in fewer than 5 percent of US layoffs in 2025. The firm suspects companies are “dressing up” routine cost cuts as AI transformation because it plays better with investors.

What many executives are actually doing is simpler than the rhetoric suggests. They are using AI as strategic cover for cost restructuring — cutting headcount not because AI is delivering productivity gains, but because they need to free up capital to invest in AI, and because “AI-driven transformation” is a better story than “margin pressure” in an analyst meeting. This is rational behavior, not cynicism. But it creates a haze between preparing for AI and benefiting from it. Organizations that mistake one for the other will over-commit before their foundations are ready.

So what does deliberate preparation actually look like?

First, what it should not look like: elaborate programs, heavy governance structures, multi-year roadmaps with dozens of workstreams. McKinsey’s own researchers warn of organizations ending up with “more pilots than Lufthansa” — AI everywhere except in the profit-and-loss statement. The complexity of this challenge requires extreme simplicity in the approach. What works is a small number of clear choices, executed with discipline and very few moving parts.

Start from Value, Not Technology

The first instinct most organizations have is to ask “what can AI do?” The better question is “where would better decisions, faster processes, or sharper insights change our results?”

AI opportunities are not equal. Some reduce cost. Some accelerate growth. Some open entirely new ways of competing. McKinsey’s research shows that companies seeing the most value from AI often set growth or innovation as objectives — not just efficiency. Most organizations default to cost plays because they are measurable and safe. But growth and innovation opportunities are where the disproportionate returns are — and the only strategic insurance against disruptive newcomers. Ignoring them is how you optimize a business into irrelevance.

Practically, this does not require months of analysis. A couple of well-designed leadership-sponsored workshops yielding a one-page map of where AI fits across cost, growth, and innovation is enough to create focus. Customer service, back-office processes, and parts of supply chain are areas where standing still already means falling behind. Others depend on foundations that are not yet in place. Making that distinction visible keeps the plan honest.

In one engagement, a multinational consumer group asked a precise question: can AI turn commercial knowledge into growth while managing risk? We designed a multi-agent sales assistant — with agents handling order recommendations, revenue optimization, and policy enforcement. It achieved over 95% precision versus expert advisors, 65%+ adoption, and a significant lift in sales productivity, touching revenue in the hundreds of millions. The technology mattered. But what mattered more was starting from a business question.

Accelerate Architecture Modernization

AI deployment spans a wide range — from giving teams a Copilot to embedding intelligent agents deep inside business processes. These require very different levels of technical investment. Conflating them is a common mistake.

Even a Copilot rollout is not trivial. Deciding which documents and knowledge sources ground it — and which do not — is a real governance challenge. Get it wrong and you either expose confidential material or produce generic outputs that nobody trusts. But the alternative is worse: people quietly using consumer AI tools with no guardrails, feeding proprietary information into public models.

The deeper challenge is process architecture. The real barrier to AI at scale is not data alone — it is the ability to blend human judgment, AI agents, and data-driven insights inside day-to-day work. McKinsey describes this as moving humans “above the loop” — supervising AI-driven workflows rather than executing tasks within them. That shift demands flexible, modular platforms. Most organizations do not yet have them.

The practical approach is not to modernize everything at once. Identify the one or two integration points where rigidity is actually blocking AI, and start there. Accept tactical workarounds where the learning justifies some technical debt — but register it and use the learning to prioritize structural modernization. Temporary shortcuts that become permanent wiring are exactly how complexity becomes gridlock.

Activate the Organization

Strategy and architecture do not execute themselves. People do. The organizational side of AI is where most efforts stall.

It starts at the top. AI is a change effort, not a technology project. If the senior team is not visibly committed — setting direction, removing obstacles, orchestrating investments — the rest of the organization will read the signal and wait. Without executive sponsorship, transformation dies in the middle of the org chart.

But commitment at the top does not automatically reach mid-level managers — directors, process owners, team leads — where AI either gets woven into real work or stalls in pilot mode. These people are usually rewarded for stability, not experimentation. If you want them to change how their teams work, change the incentives. Make experimentation rational, not heroic. In areas of strategic importance, consider ring-fenced teams whose job is to experiment — with lower operational pressure and clear goals.

McKinsey’s research suggests 75 percent of roles will need fundamental reshaping in the next two to three years — including people leading teams and those who report to them. Nearly half of leaders already see skill gaps. Have an honest conversation across the organization about what AI means — the opportunities and the disruptions. The real talent threat is not AI replacing your people. It is your competitors’ people becoming dramatically more capable than yours. Offer training that connects to actual roles and real career paths — not a generic AI awareness program, but targeted skill building that helps people see themselves in the future you are describing.

Make Change Possible

Most organizations already have change processes — planning cycles, decision gates, business cases. The problem is that these mechanisms were designed for predictable, large-scale programs. AI does not work that way. Some initiatives fail fast. Others surprise. Governance needs to keep pace without suffocating progress.

One approach that works: set aside incubation investment — a defined pool of funding with lightweight access, designed to get qualifying initiatives moving quickly. Reward not just results but discipline. When people see that following the guardrails is the fastest path to resources, the guardrails reinforce themselves.

What matters most is a simple rhythm: regular review, rebalancing, and honest assessment of what is working. Not a transformation office with a 20-person staff. The value map informs where to invest in architecture. The architecture reveals what is limiting speed. The organizations that govern these as a connected system, with the lightest mechanism that works, are the ones that turn experiments into lasting advantage.

How to Start

You do not need to solve everything at once. Start with an honest view of where you are: what is ready, what is not, where the real opportunities sit, and how fast the organization can absorb change. Build a plan that balances ambition with realism, sequenced so early moves build confidence and prepare the ground for what follows.

As I argued in October — and as growing evidence confirms — the winners in AI will not be the companies running the most experiments or the most elaborate programs. They will be the ones that build the simplest system capable of learning and adjusting.

I work with leadership teams on exactly this — translating AI ambition into a practical roadmap that aligns where you want to go with what it takes to get there. If you are working through these questions, I would welcome the conversation.

References

Some articles may require paid subscriptions.

The State of AI in 2025: Agents, Innovation, and Transformation
McKinsey & Company, 2025
https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

The Agentic Organization: Contours of the Next Paradigm for the AI Era
McKinsey & Company, 2025
https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/the-agentic-organization

AI Is Everywhere. The Agentic Organization Isn’t—Yet
McKinsey & Company, 2026
https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/ai-is-everywhere-the-agentic-organization-isnt-yet

Five Trends in AI and Data Science for 2026
MIT Sloan Management Review, 2026
https://sloanreview.mit.edu/article/five-trends-in-ai-and-data-science-for-2026

Firm Data on AI
NBER Working Paper 34836, 2026
https://www.nber.org/papers/w34836

Why AI Won’t Wipe Out White-Collar Jobs
The Economist, 2026
https://www.economist.com/finance-and-economics/2026/01/26/why-ai-wont-wipe-out-white-collar-jobs

The Tech Jobs Bust Is Real. Don’t Blame AI (Yet)
The Economist, 2026
https://www.economist.com/finance-and-economics/2026/04/13/the-tech-jobs-bust-is-real-dont-blame-ai-yet

The Hard Job of Driving Value from AI

| Charlie Ranalli

Generative AI arrived quickly, with bold promises dominating headlines. But anyone who has lived through past technology shifts knows the real story: the hard job of driving value is not proving it can work – it is adopting it to transform how companies work. Just as with PCs, e-commerce, or mobile, impact will not be fast or easy.

Customers, colleagues, and friends — nearly everyone I exchange perspectives with is already using GenAI to work faster, smarter, and better. That matches findings from a recent paper by Andrew McAfee and colleagues, which shows the technology delivering rapid productivity gains across many occupations. (And yes, I used GenAI tools in researching this article.)

The challenge is that individual gains – especially in knowledge-based work – don’t automatically scale into business or economic impact, at least in most sectors. To matter, they must flow into optimized processes, unit economics, and value levers like prices, margins, or productivity in high-volume workloads. As McAfee and his co-authors note, that means “complementary innovations and organizational reinvention.”

The good news is we finally have some evidence to work with. Early experiments show where AI can create traction and where it tends to stall. They also make one thing clear: getting value out of AI means tough choices and visible commitment from the top team.

The Road to Structural Gains

Across industries, AI pilots are everywhere – copilots for coders, chatbots for service, tools to automate routine tasks. They show what’s possible, but most remain stuck at the pilot stage. Local productivity improves, yet the broader structure of how work gets done stays the same.

This is familiar. Every major technology shift follows a pattern: early enthusiasm, a rush of experimentation, and only later the hard work of embedding new tools into the fabric of business.

The PC revolution is a widely researched case. Email, spreadsheets, and collaboration tools became available in the 1980s. Yet measurable productivity gains only showed up in the mid-1990s, once workflows were redesigned, people retrained, and cultures changed. It took 15 years before the technology showed up in macroeconomic statistics.

E-commerce followed a similar path. Despite 30 years of innovation, investment, and exceptional consumer convenience, only about 16% of U.S. retail spending is online today.

Gartner’s 2025 Hype Cycle places generative AI just past the Peak of Inflated Expectations, signaling that today’s surge of pilots and proofs of concept will only translate into value once organizations take on the harder work of reinvention.

Generative AI is best understood as a general-purpose technology, much like electricity, PCs, or the internet: broad in potential, but dependent on complementary change before its impact can be fully realized. Research by Andrew McAfee and colleagues suggests this wave could diffuse faster than earlier ones because the infrastructure is already in place, tools are widely accessible, and adoption often requires less retraining. Yet the speed of progress still depends heavily on an organization’s own maturity – some have ways of working and modern tech stacks ready to absorb AI, while others face years of groundwork before they can capture meaningful value.

Early signals remain ambiguous. A Stanford study found that workers aged 22–25 in AI-exposed roles saw employment fall by about 13% since 2022, while more experienced peers held steady – suggesting the first effects of AI might be showing up in hiring rather than productivity. A more recent analysis from Yale’s Budget Lab offers a different lens: instead of losses, it examines shifts in the occupational mix – which tasks are being automated, which are being augmented, and how work itself is changing as AI spreads. Apparent declines in some roles may instead reflect a reallocation of tasks. The evidence remains faint — more about changing work than vanishing jobs.

As a technology executive and entrepreneur who lived through the dot-com wave, the rise of e-commerce, mobile, and social media, I’ve seen this pattern repeat: first the hype, a dose of disillusionment, and finally the long, grinding work of structural change. Generative AI may spread faster than past waves, but the hard job to generate value will be the same.

Why AI Value Will Be Slow to Materialize

Integration Matters More than Deployment

Customer service bots, document summarizers, and coding copilots are easy to deploy. They spread fast but don’t transform most industries. The next step is Agentic AI – systems that combine business logic and different types of AI to perform multi-step workflows, interact with humans, and span business functions.

The challenge is integration. Agents don’t operate in isolation. They need redesigned interfaces and workflows: decision points, exception handling, escalation rules, and system integration. As McKinsey observed after studying 50+ agentic AI builds: “It’s not about the agent, it’s about the workflow.”

“Fewer than 10% of generative AI use cases have made it past the pilot stage.”
“Only 21% of organizations using gen AI report that they have fundamentally redesigned at least some workflows.”
“Among the 25 organizational practices tested, workflow redesign has the strongest relationship with reported EBIT impact.”
– McKinsey, State of AI: How Organizations Are Rewiring to Capture Value (2025)

Reengineering Operating Models

Some industries will see direct disruption – translation, for example, where AI can deliver acceptable quality at near-zero cost. But most require operating model reinvention: reconfiguring how decisions are made, how accountability flows, and how value is created.

McKinsey’s State of AI 2025 survey illustrates the gap: “Fewer than 10% of generative AI use cases have made it past the pilot stage.” Tools are spreading, but few firms have revamped their operating models to absorb them.

No Sustainable Edge in Base Technology

Generative AI is trained on public domain knowledge and released broadly. As MIT Sloan has argued, AI itself is unlikely to provide sustainable competitive advantage.

McKinsey’s survey is blunt: “Only 21% of organizations using gen AI report that they have fundamentally redesigned at least some workflows. Among the 25 organizational practices tested, workflow redesign has the strongest relationship with reported EBIT impact.”

The edge comes from differentiated data, proprietary processes, and organizational creativity. Cloud platforms can be acquired easily; unique ways of working cannot.

The Human Factor

AI changes organizational dynamics. Vertical supervising tasks give way to orchestrating systems of people and agents. Leaders must re-skill employees, manage resistance, and build trust in AI-driven systems.

Human-in-the-loop design is critical. Generative AI introduces risks of hallucinations, bias, and IP leakage that can’t be solved by technology alone. In a Harvard Business Review article, Andrew McAfee and his co-authors argue that employees need to be trained to recognize these risks, escalate them, and build confidence in using AI responsibly.

And it’s not only frontline or entry-level roles. A recent MIT Sloan article argues that AI will increasingly automate coordination and monitoring – traditional managerial functions – while creating demand for new roles in oversight, design, and orchestration.

Adoption will be uneven. Early enthusiasts embrace new tools; skeptics push back, especially after errors. Clear accountability, transparent communication, and cultural adaptation are as critical as technical readiness.

Platform Readiness: Data, Systems, and Governance

Finally, structural adoption depends on platform readiness. AI requires clean, governed data — still a rarity in most organizations. Just as critical is the underlying architecture: many ERP, CRM, and supply chain systems are built on rigid interfaces and hard-coded business logic. To take advantage of AI, those systems need more flexible designs that can support dynamic workflows. AI can’t be simply bolted onto legacy systems.

Governance must also mature alongside technology. Privacy, compliance, and risk management can’t wait until systems are scaled if they are exposed to the outside world – they need to be embedded from the start.

The Productivity Promise Is Still Real

Despite these challenges, the long-term upside is enormous. McKinsey estimates that Agentic AI could unlock $450–650 billion annually by 2030 in advanced industries such as manufacturing, logistics, and energy.

Functions like supply chain, software development, and customer service can all be reimagined. But because the base technology is broadly available, the competition to innovate will be fierce. The winners will be those that integrate faster and deeper, building on proprietary assets.

From my experience with past technology shifts, this is where the champions pull ahead. When mobile and social networks disrupted consumer engagement, the companies that succeeded weren’t the ones with the flashiest apps – they were the ones that cleverly reimagined marketing, sales, and service for two-way, mobile-first interactions. The same will be true with AI.

From Experimentation to Transformation

The practices below reflect what I’ve seen in working with clients, along with insights from global leaders and academics. Together, they highlight what helps organizations move beyond experimentation toward real transformation.

Activate the Leadership Team

Transformation begins at the top. This is not just a technology project but a significant change effort. The senior team has the credibility to ask the hard questions, the clout to mobilize resources quickly, and the leverage to remove obstacles. If they aren’t in the room and visibly committed, the rest of the organization will sense it – and resistance will dampen progress.

Map Opportunities

With customer teams, we start broad and deep. The goal isn’t just to identify obvious efficiencies, but to uncover disruptive opportunities. We map ideas across two dimensions: strategic business pillars, and the angles of innovation that AI can unlock.

This stage is about thinking differently – exploring what could change the rules of the game rather than just automate today’s tasks. It’s a step for conceiving opportunities, not a commitment to chase them all.

Invest in a Balanced Portfolio

Once opportunities are conceived, the next step is to commit to a portfolio that balances ambition with pragmatism. In practice, the portfolios that resonate most with leadership teams usually contain three kinds of bets: disruptive ideas that carry risk but could redefine value creation; quick wins that deliver visible results and momentum; and defensive moves that may not create lasting advantage but are quickly becoming industry baseline.

Start with the Questions

AI is not an end in itself. The right starting point is a business question: If I could forecast X more accurately, what would change? If I could automate Y, how much capacity would I free up?

One industrial client is embedding AI agents into a reactive B2B service chatbot. The initial aim is faster response, but the next goal is to extend into proactive commercial processes – identifying cross-sell opportunities, guiding orders, anticipating needs. The key question wasn’t an open-ended “what can AI do,” but “can AI turn knowledge into growth while managing risk up front?”

Advance Initiatives and Capabilities

Run promising use cases first but do so with discipline: set substantial business goals, track adoption and value, and be ready to kill experiments that don’t deliver. In parallel, invest in the capabilities that enable scaling: data assets, governance, orchestration technology, and workforce skills.

One of my retail clients launched a Copilot program to organize and classify knowledge. The first phase focuses on individual and small team productivity, but the explicit goal is to mature into AI-enabled collaborative innovation. Sequencing matters: early initiatives build experience and momentum, while capabilities ensure the organization is ready to accelerate what works.

Close Business, Technology, and Risk Collaboration

AI can’t sit in a technical silo. Business leaders must co-own initiatives. To build trust and ensure each step forward stands on firmer ground, risk and compliance need to be involved from the start. Governance has to be continuous, with clear paths for when agents fail.

One useful approach, highlighted in HBR and practiced by most early adopters, is to move deliberately in stages: experiment in a sandbox, then run tightly scoped pilots, and only scale once risk and governance structures are proven to hold.

The Case for Simplicity

When teams start mapping AI opportunities, the challenge isn’t scarcity – it’s abundance. The possibilities are endless, and the actions required to pursue them quickly become complex and interdependent. Without a simple top-level frame, discussions can fragment across business, operations and technology perspectives.

A practical way to bring structure is to group AI applications into broad categories, for example:

Predictive & Optimization – improving foresight and decision quality through analytics and modeling.

Autonomous Decision & Action – embedding intelligence into processes that can sense and act in real time.

Knowledge-Management & Generative – expertise, content, and amplifying human capability.

End-to-End – connecting and orchestrating workflows across functions to create compound value.

Illustrative opportunity matrix. Sanitized from project material to show structure, not content.

This structure, borrowed from a recent workshop, isn’t meant to be followed verbatim – it must be reframed based on each organization’s industry and AI posture, and paired with strategic pillars to give a full 360-degree view. It simply shows how to organize a portfolio around general concepts rather than specific technologies or use cases.

The Takeaway

AI opens unprecedented opportunities but is not a silver bullet. Pilots and use cases are necessary, but they are not sufficient. Like PCs, e-commerce, and mobile before it, AI’s real value will only come when organizations reinvent their structures to embed it deeply in business and operating models.

The winners won’t be those who run the most experiments. They will be the organizations willing to rewire themselves so that AI becomes part of the fabric of how business is done. For leaders, that means rolling up your sleeves and doing the hard job of transformation.

References

The impact of generative AI as a general-purpose technology
MIT Sloan School of Management, 2024
https://mitsloan.mit.edu/ideas-made-to-matter/impact-generative-ai-a-general-purpose-technology

Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence
Stanford Digital Economy Lab, 2025
https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine/

Evaluating the Impact of AI on the Labor Market: The Current State of Affairs
Yale Budget Lab, Yale University, 2025
https://budgetlab.yale.edu/research/evaluating-impact-ai-labor-market-current-state-affairs

One Year of Agentic AI: Six Lessons from the People Doing the Work
McKinsey & Company, 2024
https://www.mckinsey.com/capabilities/quantumblack/our-insights/one-year-of-agentic-ai-six-lessons-from-the-people-doing-the-work

The State of AI: How Organizations Are Rewiring to Capture Value
McKinsey & Company, 2025
https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

Why AI Will Not Provide Sustainable Competitive Advantage
MIT Sloan Management Review, 2025
https://sloanreview.mit.edu/article/why-ai-will-not-provide-sustainable-competitive-advantage/

How to Capitalize on Generative AI
Harvard Business Review, 2023
https://hbr.org/2023/11/how-to-capitalize-on-generative-ai

Why Robots Will Displace Managers – and Create Other Jobs
MIT Sloan Management Review, 2025
https://sloanreview.mit.edu/article/why-robots-will-displace-managers-and-create-other-jobs/

The Economic Potential of Generative AI: The Next Productivity Frontier
McKinsey & Company, 2023
https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier