AI Academy · Book
Executives & Directors · Module 01 · Chapter 001

AI Is Changing Everything

It is, but not on its own. Nearly every organization now uses AI, and few get much from it. This chapter covers what is actually changing, what the evidence says, and why the same technology produces very different results.

≈ 15 min read

After this chapter you can

  • Explain what "AI is changing everything" does and does not mean, with evidence.
  • Describe how AI enters the three flows of work, knowledge and decisions.
  • Distinguish automation, augmentation and transformation, and test which one an initiative really is.
  • Explain why the same technology produces very different results.
  • Hold a stance of urgency without panic, including what the law already expects.

In 1900, American factories had electricity and almost nothing to show for it. Central power stations had been running since the early 1880s, and electric motors were for sale. Yet electric motors supplied less than 5 percent of the mechanical drive in US manufacturing. Most factory owners who did buy a motor bolted it onto the system they already had: one large power source turning a long line shaft, with belts and pulleys running to every machine. The new technology went in, the factory stayed the same, and so did productivity.

The payoff came in the 1920s, close to forty years after the first power stations. It came when a new generation of managers stopped treating electricity as a better steam engine. They put a small motor on each machine, laid out the floor by the logic of the work, and built lighter, single-story plants that the old shafts had made impossible. The economist Paul David showed that this redesign, not the arrival of the dynamo, is what drove the gains1.

Hold that picture. It is one of the most useful ideas for an executive trying to make sense of artificial intelligence.

What “everything” means, and what it does not

The phrase AI is changing everything invites suspicion, and it should. It does not mean every process will be automated, or that every function will change at the same speed. It means three narrower things, each of which is supported by evidence.

AI has left the lab. In McKinsey’s 2025 global survey, 88 percent of organizations reported using AI in at least one business function2. It is no longer confined to a data-science team or an innovation unit.

It reaches wherever knowledge work happens. Today’s AI is very good at the raw material of office work: language, documents, images, code and patterns in data. That raw material runs through finance, people management, legal, sales, operations, customer service and engineering alike. Few functions have nothing for it to touch. This is a statement about capability, not results: being able to reach a function is not the same as paying off in it, which is the third point.

The value is very unevenly spread. The same survey found that only about a third of organizations had begun to scale AI beyond pilots. Only around 6 percent were “high performers” that attribute 5 percent or more of their EBIT to AI and describe the impact as significant2. McKinsey’s 2026 survey found that share essentially unchanged3. The three figures measure different things. “Using AI” means at least one function has it; “scaling” is about how far it has spread; “high performer” is about money. All three are shares of survey respondents, not of all companies. Almost everyone has the technology. Very few have changed the business.

The third point is the one this book returns to most often. It is the factory story again.

AI at the center, connected to eight business functions from strategy to engineering.AIinside the workStrategyFinancePeopleLegalSalesOperationsServiceEngineering
Figure 1.1.1 AI sits inside the enterprise, not beside it. Every function handles the language, documents and data it works on.

What AI amplifies

Every general-purpose technology amplified something specific. Steam and electricity amplified physical work. Computers amplified calculation. The internet amplified communication and distribution. AI amplifies knowledge work: drafting, summarizing, searching, classifying, predicting and checking. The next chapter places this in the longer history of industrial revolutions. For now, the consequence is what matters. Knowledge work is the core of most modern organizations, so AI is not an IT topic that some functions can ignore. Finance wants better forecasts. HR wants faster and fairer hiring. Legal wants document review without losing judgment. Service wants answers that are fast, accurate and consistent. Each of these is a different problem, and each has a reason to care.

Three flows run through every organization

A useful way to see the change is to follow what moves through an organization in an ordinary week. Reports are written, contracts reviewed, customers answered, forecasts built, code shipped, policies interpreted. Underneath all of it, three things are flowing.

  • Work: someone performs a task.
  • Knowledge: someone finds, writes down or remembers information.
  • Decisions: someone chooses what happens next.

Earlier software usually changed one process inside one flow: a new finance system changed how invoices move. AI enters all three flows at once, which is why it feels larger than a new application.

A table of the three flows, showing what AI now does in each and what stays human.FlowWhat AI now doesWhat stays humanWorkDrafts, classifies, routes and checksroutine outputJudging quality and handling the exceptionsKnowledgeFinds, summarizes and connects informationacross sourcesKnowing which source to trust, and whenDecisionsPrepares options, evidence and afirst recommendationChoosing, and answering for the choice
Figure 1.1.2 AI changes how work, knowledge and decisions are prepared. Accountability for the decision does not move.

The right-hand column matters as much as the left. The preparation of a decision is changing fast. Responsibility for the decision is not. An executive who keeps that distinction clear can adopt AI aggressively without giving away accountability.

What the evidence says about individual work

Claims about AI productivity are cheap, so it is worth knowing the handful of careful studies that measure it. Three controlled studies stand out, and together they give a more useful picture than any vendor benchmark.

In a study of more than 5,000 customer-support agents, an AI assistant raised issues resolved per hour by 15 percent, with the gains going mostly to less experienced agents4. A controlled writing experiment found AI cut task time by 40 percent and raised rated quality by 18 percent5. Those two figures are not on the same scale, so they should not be compared directly. The study every executive should know is a third one. Researchers gave 758 consultants at a global consulting firm realistic tasks, with and without AI6. On tasks inside what the authors call the jagged frontier of AI capability, consultants with AI finished 12.2 percent more tasks and worked 25.1 percent faster. Their output was rated more than 40 percent higher in quality. On a task that looked similar but fell outside the frontier, consultants using AI were 19 percentage points less likely to reach the correct answer than those working without it. Note the units: the quality gain is a percentage of a rating, while the loss is a difference in the share of correct answers.

Inside the frontier AI made consultants faster and better; outside it, they became more likely to be wrong.INSIDE THE FRONTIERTasks the AI handles well25% faster, 40% higher qualityOUTSIDE THE FRONTIERSimilar-looking tasks itgets wrong19 points less likely to be rightvs
Figure 1.1.3 The same tool helped on one task and hurt on another (Dell’Acqua et al., 2023).

Three management lessons follow, and none of them is about technology.

  1. The gains are real but uneven. They are largest where work is repetitive, language-heavy and well defined. They are smallest, and can turn negative, for experts doing work the system does not handle well.
  2. The frontier is invisible from the outside. AI fails without warning. People cannot see where the frontier lies by looking at the task. Organizations therefore need deliberate checking, not blind trust or blanket bans.
  3. AI can level up the newest people fastest. In the support study the gains went mostly to newer agents. That changes how you onboard, coach and staff a team. It is a reason to rethink the work, not just to add a tool to it.

Three levels of change

Not all AI initiatives are the same kind of change. Separating them is the most useful distinction in this chapter, because each level asks something different of leaders.

Three ascending steps from automation to augmentation to transformation; advantage sits at the top.AutomationRepeat the taskAugmentationAssist the humanWHERE MANY ARE TODAYTransformationRedesign the work
Figure 1.1.4 Three levels of change. Many organizations sit in the middle; advantage is created at the top.

Automation means a machine performs a repetitive task end to end, such as capturing invoice data or resetting passwords. The goal is usually cost, speed or fewer errors. It is valuable, and it is the oldest of the three.

Augmentation means a person stays in control while AI raises the speed or quality of their work. A lawyer reviews a first-pass contract summary, an engineer works with a coding assistant, a manager walks into a meeting already briefed. The person still owns the outcome. AI changes the effort needed to get there. Much of organizations’ use of AI today appears to sit at this level. That is a reading of the survey pattern, wide adoption and rare scaled value, not a measured share.

Transformation means the organization redesigns the work around what AI makes possible. The process changes, and so do roles, controls, measures and often the customer’s experience. Picture a claims function that no longer starts with a person reading every document. AI assembles the case, specialists handle the exceptions, and managers coach a different kind of team. That is the electric factory, not the motor bolted onto the line shaft.

The labels are easy to agree with and easy to misapply. Most initiatives described as transformation are augmentation with a larger budget. The table below gives a quick test.

A table comparing automation, augmentation and transformation on what changes, who owns the outcome, what is measured, and an example.AskAutomationAugmentationTransformationWhat changes?One task, end to endOne person's effortProcess, roles,controls, measuresWho ownsthe outcome?The process ownerThe same person as beforeA redesigned teamWhat do you measure?Cost, errors, cycle timeOutput and quality per personBusiness result: growth,service, riskExampleCapturing invoice dataA lawyer checks an AI summaryClaims built around anAI-assembled case
Figure 1.1.5 A quick test: if roles, measures and controls are unchanged, it is not transformation, whatever the program is called.

None of the three levels is wrong. A sensible portfolio has all three. The mistake is to fund augmentation and expect transformation’s results, or to call a tool rollout a transformation because the budget was large.

Same technology, different outcomes

If nearly every organization uses AI, why do so few get significant value from it? The evidence points to the same answer the factories gave a century ago.

A timeline from the first power stations in the 1880s to the productivity surge of the 1920s, when factories were redesigned.1880sPower stationsopenElectricity is available1900Under 5% offactory driveMotors bolted ontoold shafts1919About half of driveAdoption is wide; gainsare not1920sProductivity surgesFactories redesignedaround motors
Figure 1.1.6 Electricity paid off only when the factory was redesigned around it, roughly forty years on (David, 1990).

McKinsey’s surveys show the modern version of the same gap. Adoption is close to universal. Scaled use and financial impact are not.

Bars falling from 88 percent using AI to 6 percent attributing at least 5 percent of EBIT to it.Use AI in at leastone function88%Report any EBIT impactfrom AI39%Have begun to scale AI33%High performers: 5%+ ofEBIT from AI6%Source: McKinsey global survey, 1,993 respondents · 2025
Figure 1.1.7 Nearly everyone uses AI; about 6 percent get significant financial value from it. The 2026 survey found the same pattern.

What separates the small group at the end of that chart? McKinsey’s analysis is clear. High performers were nearly three times as likely as other organizations to have fundamentally redesigned their workflows. They also set transformation goals rather than efficiency goals alone, and their senior leaders were more likely to show visible ownership of the effort2. Their technology is not different. What they do with it is.

Take two insurers that license the same platform. One connects a few systems, gives claims handlers a login and keeps the old process, targets and staffing. The other picks one painful process, from first notice of loss to settlement, lets AI assemble the case file, retrains managers to coach exception handling and samples quality every week. Only the second changes how claims move. The difference is not the vendor or the budget. It is the set of leadership decisions around the technology. Which problem to pick, how the work changes, how people are trained, what is measured, and who is accountable. Those decisions cannot be bought, and they cannot be delegated to the technology team.

Why this lands on every executive’s desk

It is tempting to treat AI as the technology function’s business, with business leaders as customers. That framing fails for three reasons.

The value depends on business decisions. Choosing the problem, redesigning the process, changing roles and measures, and leading people through it are leadership work. Technology teams are essential to every one of these. They cannot own them.

The workforce questions are yours. If AI levels up newer employees and shifts what experts spend their time on, then hiring, career paths, team design and performance measures all change. No platform settles those questions.

The law already treats organizations that use AI as responsible for it. You do not have to build an AI system to carry obligations for it.

None of this means every executive must become technical. It means every executive must be able to ask the right questions, tell real opportunity from theater, fund the right work and refuse the wrong work.

Urgency without panic

There are two easy mistakes. One is complacency: AI is optional, so wait and see. The other is panic: buy something large now and look for the value later. Both are common, and both are expensive.

A spectrum from complacency to panic, with deliberate learning in the middle.ComplacencyWait until it settlesPanicBuy now, find value laterDeliberate learningReal problems, fast evidence
Figure 1.1.8 The right stance sits between the two failures. Learn fast on real problems, and fund only what proves its value.

The case for urgency is not that the technology will run away from you. It is that the scarce asset is organizational learning, and learning compounds. Each real deployment teaches you how your data behaves, which processes are ready, how people adopt new tools, and which controls you need. Waiting does reduce some technology risk. It also delays every one of those lessons. The electric factories did not win by owning more motors. They won because their managers learned sooner how to rebuild the floor.

The case against panic is the same evidence. Spending is not learning. A large program spread across every function at once tends to produce many pilots and little redesign. That is exactly the profile of the organizations at the wide end of the chart above.

A working mental model: the fast junior analyst

For most organizations today, a good mental model for AI is an exceptionally fast junior analyst. It can read a pile of documents in minutes, draft a first version, summarize a meeting, suggest options and brief you before you walk in. That is genuinely valuable.

It is also incomplete in ways that matter. The analyst does not own the decision. It does not know your risk appetite unless you tell it. It can be confidently wrong, especially just past the jagged frontier, and it does not flag when that happens. It can expose information that should never have left the room. The executive’s role is the manager’s: set the task, check the work, decide, and answer for the outcome. AI accelerates preparation. It does not replace judgment. Later chapters refine this model when they reach AI agents that act as well as advise.

What to do this quarter

These questions do not require a budget decision, a vendor choice or technical knowledge. They require attention. They move an organization from having AI toward changing how it works.

Check yourself

  1. Most organizations that use AI already get significant financial value from it.
  2. Giving people an AI assistant can make some of their work worse, not better.
  3. Two similar companies using the same AI tools will usually see similar results.
  4. AI mainly matters to technology companies.
  5. The less experienced people on a team often gain the most from AI assistance.
  6. Under the EU AI Act, only the companies that build AI systems have obligations.

Reflection: name one process

What comes next

If AI brings a change on the scale of electricity, it should not be treated as a surprise. Leaders have lived through transformations like this before, and the factory story is one of them. The next chapter, The Four Industrial Revolutions, asks what makes this one different: its speed, its scope, and the kind of work it amplifies.

Laws referenced

Not legal advice. Laws change; verify before relying on this, and consult counsel for decisions.

EU AI Act · EU

Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744

Risk-based rules. Prohibited practices include social scoring, untargeted scraping of facial images, and emotion recognition in workplaces and schools (with narrow exceptions). High-risk systems (Annex III: biometrics, safety components of critical infrastructure such as energy, water and traffic, employment and worker management, credit, education, essential services, law enforcement, migration, justice) need risk management, data governance, documentation, logging, human oversight, human oversight that keeps people able to understand the system, notice automation bias (over-reliance on its output), override it or stop it (Art. 14(4)), appropriate accuracy, robustness and cybersecurity (Art. 15), automatic logging of events (Art. 12), a provider quality-management system (Art. 17) and conformity assessment. An Annex III system is not high-risk if it poses no significant risk of harm, for example a narrow procedural or preparatory task that does not replace human assessment; systems that profile people are always high-risk, and a provider relying on this exception must document it and register (Art. 6(3)). Deployers of high-risk AI must use it as instructed, assign competent human oversight, monitor its operation, keep logs for at least six months and report serious incidents (Art. 26); employers must inform workers' representatives (Art. 26(7)). Public bodies, private providers of public services, and deployers of credit-scoring or life and health insurance pricing systems must carry out a fundamental-rights impact assessment before first use (Art. 27). Providers must run post-market monitoring (Art. 72). A deployer that puts its name on a high-risk system, substantially modifies it, or changes its purpose so that it becomes high-risk takes on the provider's obligations (Art. 25(1)). A substantial modification (Art. 3(23)) of a high-risk system needs a new conformity assessment, unless the change was pre-determined and documented at the first assessment, as with planned continuous learning (Art. 43(4)). Providers of general-purpose AI models (from 2 Aug 2025) must keep technical documentation, have a policy to comply with EU copyright law including text-and-data-mining opt-outs, and publish a sufficiently detailed summary of training content (Art. 53). Research, testing and development before a system is placed on the market or put into service is outside the Act, except testing in real-world conditions (Art. 2(8)). Since the 2026 Omnibus, the Art. 4 AI-literacy duty is an obligation of effort (take measures to support literacy), not of result. Fines reach EUR 35 million or 7% of global turnover for prohibited practices.

  • 2024-08-01 — Entered into force
  • 2025-02-02 — Prohibited practices (Art. 5) and the AI-literacy duty (Art. 4) apply
  • 2026-07-27 — Omnibus softens Art. 4: providers and deployers must take measures to support AI literacy; no specific level must be guaranteed
  • 2025-08-02 — General-purpose AI model obligations apply; governance and penalties regime in place
  • 2026-08-02 — Transparency duties (Art. 50) apply: disclose AI interaction, label synthetic and deepfake content (marking for generative systems already on the market: 2 Dec 2026)
  • 2027-12-02 — High-risk obligations for Annex III systems (e.g. hiring, credit, education, essential services) - moved from 2 Aug 2026 by the 2026 Omnibus
  • 2028-08-02 — High-risk obligations for AI in products regulated under Annex I

Last verified 2026-10-06 · official text

US federal AI policy and state-law preemption push · US - federal

Executive Order 14365 (11 Dec 2025); White House National Policy Framework for AI (20 Mar 2026)

The federal government is seeking to override "burdensome" state AI laws and has asked Congress for a national framework. These are executive actions and recommendations, not a federal AI statute. State laws such as NYC LL144 still apply until changed by law or the courts.

Last verified 2026-10-06

NYC Local Law 144 (automated employment decision tools) · US - New York City

NYC Local Law 144 of 2021; DCWP rules

An automated tool that substantially assists hiring or promotion decisions needs an independent bias audit within the past year, a published summary of results, and notice to candidates at least ten business days before use. Penalties USD 500 to 1,500 per violation.

  • 2023-07-05 — Enforcement began

Last verified 2026-10-06 · official text

Colorado AI law · US - Colorado

SB 24-205, repealed and re-enacted by SB 26-189 (signed 14 May 2026)

The first US state law aimed at algorithmic discrimination in consequential decisions. It was delayed and then replaced by a narrower, notice-based framework before it ever took effect. Expect further change.

  • 2027-01-01 — Operative requirements of SB 26-189 take effect

Last verified 2026-10-06

California privacy rules on automated decisions (CCPA regulations) · US - California

California Consumer Privacy Act; CPPA regulations on ADMT, risk assessments and cybersecurity audits (approved by OAL Sept 2025)

The most concrete US privacy rule on AI. Businesses that use automated decision-making technology to make a significant decision about a California resident (finance or lending, housing, education, employment or pay, healthcare) must give notice before use, offer an opt-out unless an exception applies, and answer access requests. Processing that poses significant privacy risk needs a documented risk assessment. There is no comprehensive federal privacy statute; about 20 states have their own laws, and California's is the reference point.

  • 2026-01-01 — Updated CCPA regulations take effect; risk-assessment duty applies to new high-risk processing
  • 2027-01-01 — ADMT duties for significant decisions: pre-use notice, opt-out (with exceptions) and access (some firm alerts cite enforcement from 1 Apr 2027)
  • 2028-04-01 — Attestation of 2026-2027 risk assessments due to the CPPA; cybersecurity audits phase in 2028-2030 by revenue

Last verified 2026-10-06

References

  1. Paul A. David. The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox. American Economic Review 80(2). 1990.
  2. McKinsey & Company (QuantumBlack). The state of AI in 2025: Agents, innovation, and transformation. McKinsey & Company. 2025.
  3. McKinsey & Company (QuantumBlack). The state of AI in 2026: On the road to ROI. McKinsey & Company. 2026.
  4. Erik Brynjolfsson, Danielle Li and Lindsey Raymond. Generative AI at Work. The Quarterly Journal of Economics 140(2). 2025.
  5. Shakked Noy and Whitney Zhang. Experimental evidence on the productivity effects of generative artificial intelligence. Science 381(6654). 2023.
  6. Fabrizio Dell'Acqua et al. Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality, HBS Working Paper 24-013. Harvard Business School. 2023.

Further reading

Sources last verified 2026-10-08.